Critically ill neonates born every day require the progressively high technologic care that is available in the neonatal intensive care unit (NICU). The demand for neonatal critical care is on the rise. Halpern and Pastores1 found that neonatal intensive care beds in the United States have increased by 8% and the infant mortality rate still remains high compared with that in other developed countries.2 Specifically, the 2010 US infant mortality rate of 6.1 deaths per 1000 live births was the highest infant mortality rate among 26 developed countries included in the Organization for Economic Co-operation and Development.2 Although NICU care has improved over the past 30 years and survival rates are increasing,3 advances are needed for the improvement of critical care for the most vulnerable newborns.

Clearly decreasing the infant mortality rate is a critical goal; however, neonatal survival is no longer the ultimate goal of neonatal intensive care. Vulnerable neonatal populations, specifically the youngest and smallest babies, are at greatest risk of mortality as well as a lifetime of morbidity following a NICU admission.4,5 Researchers and clinicians are working diligently to optimize care for these critically ill neonates and thereby decrease the risk of mortality and reduce the risk of lifetime morbidity. Utilization of a tool to operationalize the degree of illness severity for neonates has potential for use as a bedside clinical tool as well as a research tool. Current trends utilizing electronic medical records in the NICU could allow for automation of an illness severity score calculation by using computer code to pull chart data to assign an illness severity score. Operationalizing illness severity scores based on physiologic data points has the potential to help the bedside clinician establish risk at birth and monitor illness severity throughout the patient's admission. It is also important to consider the role of illness severity scores in neonatal research. An illness severity score has the potential to assist researchers in accounting for the level of illness severity at birth and to account for their illness severity when they evaluate for the effectiveness of the interventions. Clearly well-designed research studies are needed to generate new knowledge and illness severity scores have the potential to improve the data that are obtained, thereby helping researchers continue to improve the care that is provided for these critically ill newborns.

When designing research studies for this vulnerable population, it is important to consider the concept of neonatal illness severity. Neonatal severity of illness is a concept that is 2-fold: (1) determination of how acutely ill or physiologically unstable the newborn is at that time and (2) assisting in the prediction of future risk of morbidity and mortality for the neonate because of this initial risk.6

DERIVATION OF THE SCORE FOR NEONATAL ACUTE PHYSIOLOGY

One of the most common research instruments available to measure the concept of neonatal severity of illness is the revised version of the Score for Neonatal Acute Physiology (SNAP-II). The original SNAP was developed by Richardson et al.7 One of the greatest reasons for this tool was the need to accurately compare outcomes within and among NICUs. Richardson and colleagues noted that an additional source of variance in NICU cohorts is neonatal illness severity, and without controlling for this initial variation of risk at birth, outcome comparisons would be inaccurate.7 Therefore, the SNAP measurement tool was created. The SNAP was specifically designed to measure the physiologic instability of the newborn across body systems that are present in the first 24 hours of life. These physiologic measurements change overtime and as such the SNAP instrument was designed to allow for sequential measurement.7

Severity of illness is related to risk of mortality; however, there are also perinatal risk factors that influence the risk of mortality that are independent of illness severity. Therefore, the Perinatal Extension (SNAP-PE) was added to the SNAP score to quantify physiologic instability and perinatal mortality risk in one instrument.8 This instrument contains the full SNAP score and then adds in 3 additional perinatal parameters, including birth weight, small-for-gestational age, and the 5-minute Apgar score, which accounts for additional perinatal risk factors of mortality.8

The creators of the original SNAP measurement tool generated initial evidence of validity by demonstrating that increasing SNAP scores were associated with increasing mortality rates (P < .001). Additional evidence of validity of the SNAP score was demonstrated by looking for positive relationships between other variables associated with increased illness severity. For example, positive relationships were noted between SNAP scores and additional parameters, including the length of stay (R² = 0.59), nursing acuity (r = 0.59), and increased need for therapeutic interventions (r = 0.78).7 Clearly when babies are more ill, they spend more days as an inpatient in the NICU, they require more interventions in general, which includes a need for a higher level of nursing care. Finally, a positive correlation was noted between SNAP scores and the physician's estimation of mortality risk (r = 0.65).7 The SNAP-PE also demonstrated the ability to predict the combined physiologic and perinatal mortality risk in neonates. The SNAP-PE instrument was a better predictor than SNAP or birth weight alone (area under curve [AUC] ranged from 0.91 +/- 0.2 to 0.93 +/- 0.2). Furthermore, the Hosmer-Lemeshow (HL) goodness of fit test was poor for the birth weight alone but good for other models (P = .2-.99).8

The SNAP instrument had the potential to assist clinicians and researchers to quantify the concept of illness severity. However, this tool was extensive and required up to 15 minutes to evaluate many parameters including blood pressure, heart rate, respiratory rate, temperature, PO₂, PO₂/FIO₂ ratio, PCO₂, oxygenation index, hematocrit, white blood cell count, immature-to-total ratio, absolute neutrophil count, platelet count, blood urea nitrogen, creatinine, urine output, indirect bilirubin, direct bilirubin, sodium, potassium, ionized calcium, total calcium, glucose, bicarbonate, pH, seizures, apnea, and stool guaiac.6,7 The authors acknowledged that the SNAP scoring system was cumbersome and their stated goal from the initial instrument development was to eventually make the tool more parsimonious.7 However, they needed a large cohort of neonates with SNAP data to further analyze and reduce the number of parameters, while still retaining the validity of the illness severity score.

Richardson et al6 published the initial derivation and validation report of the SNAP-II and SNAPPE-II instruments from a cohort of 25 429 neonates across 30 sites from 3 neonatal networks in the United States and Canada. The revised SNAP-II instrument achieved the authors' goal of creating a more parsimonious tool that could be assessed within 2 to 4 minutes. The SNAP-II retained only 6 items and each item was weighted on the basis of the beta weight from the logistic model. The parameters and possible point values were as follows: lowest mean blood pressure, lowest temperature, PO₂/FIO₂, lowest serum pH, presence of multiple seizures, and low urine output. Just like the SNAP, the SNAP-II is a summative rating scale. The highest possible score is 115. The SNAPPE-II is also the summative rating scale as listed earlier and it adds 3 additional parameters, including birth weight, small-for-gestational age, and the 5-minute Apgar score. The highest possible SNAPPE-II score is 162. The higher the SNAP-II or SNAPPE-II score is, the more severely ill and physiologically unstable is the neonate. As described by the instrument authors, the SNAP-II was designed to measure the mortality risk due to physiologic instability and the SNAPPE-II was designed to measure the combined physiologic and perinatal mortality risk. Because perinatal factors will not change over time, the SNAPPE-II was only designed to be measured once with data from the first 12 hours following birth. However, the SNAP-II instrument was designed to assess and quantify the physiologic signs of illness that can be assessed clinically. These physiologic derangements can change over time; therefore, the SNAP-II may be useful to measure severity of illness over time.

Parsimony was a strength of this revised tool; however, equally important is that the points assigned to each parameter were empirically derived from the beta weights from the logistic model for SNAP-II.6 While some tools have been designed to measure illness severity in specific birth weights of neonates, the SNAP-II instrument was designed as a universal tool for critically ill newborns of all gestational ages and birth weights. The SNAPPE-II demonstrated good sensitivity, specificity, and HL goodness of fit across all birth weights. When stratifying the cohort via birth weight, the values were as follows: (1) across all birth weights (AUC 0.91 +/- 0.01; HL P = .9), (2) less than 1500 g (AUC 0.85 +/- 0.01; HL P = .86), and (3) 1500 g or greater (AUC 0.87 +/- 0.03; HL P = .63).6 No reliability data such as the Cronbach alpha were reported.

PURPOSE

Specifically, this literature review will synthesize all research studies that have utilized the SNAP-II instrument since its creation. As designed, this instrument measures illness severity as physiologic instability of the neonate. While most studies used this instrument to measure illness severity in the first 12 hours of life, some studies used this instrument at later time points and for sequential measurements. While the authors of the tool are clear that the SNAPPE-II is an admission score only, the original SNAP instrument was designed for sequential measurements. However, the utility of the SNAP-II at later time points and for measurements over time is unclear. This literature review seeks to describe how the SNAP-II instrument has been used across neonatal populations within the NICU. In addition, this literature review describes current evidence that is available to support the use of the SNAP-II tool at later time points and for prospective measurements of illness severity beyond the first day of life.

METHODS

All research studies that have utilized the SNAP-II instrument since it was originally derived and validated by Richardson et al6 will be reviewed. The primary author used 3 search engines to gather potential articles for review: PubMed, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Web of Science. The search terms "neonatal severity of illness" and "SNAP" were used to retrieve relevant articles. The term "SNAP" was used instead of "SNAP-II" to ensure that no relevant studies were overlooked during the initial data retrieval process. Date ranges were omitted since it was feasible to evaluate older papers to understand the full history of the SNAP-II instrument. The inclusion criteria for this comprehensive review included (1) articles available in English, (2) utilization of the SNAP-II instrument, and (3) and original research. The exclusion criteria included (1) duplication of the study from another search engine; (2) articles not available in English; (3) nonresearch articles, including review articles, poster abstracts, conference abstracts, letters to the editor, and commentaries; and (4) studies that did not utilize the SNAP-II instrument in the study.

Once studies were identified, the primary author reviewed each study and data were extracted for synthesis. A table compiled relevant data extraction and included (1) author and year of the publication, (2) study purpose, (3) study location (country), (4) sample size, (5) sample characteristics, (6) research design, (7) utilization of the SNAP-II instrument, (8) research findings, and (9) limitations and potential for bias (see Table 1). Data in the table were compiled to explore the complete utilization of the SNAP-II instrument since the revised instrument was created in 2001.

Table 1-a. Review of the relevant research articles for utilizing the score for neonatal acute physiology

RESULTS

Derivation of included studies

Studies were evaluated on the basis of the inclusion and exclusion criteria. The process of study selection is outlined later and is also depicted in Figure 1. The literature search terms "neonatal severity of illness" and "SNAP" yielded a total of 193 articles. Seventy-eight articles were duplicates and therefore eliminated. Six studies were excluded because of being published in foreign languages, including Chinese,9 Bulgarian,10 French,11 Portuguese,12,13 and Polish.14 This yielded a total of 109 abstracts for review. Review of the abstract revealed if the measurement tool for the study included the initial SNAP) or the revised SNAP (SNAP-II). Only studies that included the SNAP-II measurement tool were included; therefore, an additional 59 studies were excluded. The final step in article selection was to exclude any

nonresearch articles. Fourteen studies were excluded including 12-conference meeting or poster abstracts, one commentary, and one letter to the editor. This left a total of 36 papers for full review; however, one additional study was eliminated because the SNAP-II instrument was not used as an instrument for the study. Therefore, a total of 35 articles remained for inclusion in this review.

Characteristics of included studies

The SNAP-II is one of the most widely used measurement instruments used to operationalize the concept of neonatal illness severity. It has been used internationally in countries, including the United States,15-19 Canada,20-32 Spain,33-35 Italy,36,37 Brazil,12,38 India,39,40 France,41 Thailand,42 the Netherlands,43 Ireland,44 China,45 and Iran.46 Two studies had cohorts that were derived from populations across 2 countries, including the United States and Canada.6,47 It is important to note that of the 35 studies meeting inclusion criteria, 20 studies were conducted in Canada, the United States, or both. This includes the initial study of derivation and validation of the SNAP-II measurement tool.6

Research studies were almost equally divided between multicenter and single-center studies. Three large neonatal networks, including the Canadian Pediatric Surgery Network,20,21,24-26 Canadian Neonatal Network,6,22,27,29-32 and the Vermont Oxford Network,6,16,17,47 participated in several studies. The sample sizes varied greatly among the studies as shown in Table 2. Samples ranged from the smallest sample of 3228 to the largest cohort of 25 429.6 Nine studies included sample sizes fewer than 10015,28,33,34,36,39,42-44; however, there were 8 larger studies with sample sizes more than 1000.6,16,17,27,30-32,47 Newborns in the studies varied greatly across gestational age and birth weight. While the 2 large validation studies included newborn cohorts across the entire continuum of gestational ages and birth weights,6,47 most studies focused on preterm, late preterm, or term newborns. Some studies specifically focused on very low-birth-weight newborns19,44,48 and extremely low-gestational-age newborns16,17 which are premature populations with expected risk of illness severity. Not only was the SNAP-II used across all gestational ages and birth weights, it was also used across many specific morbidities, including congenital diaphragmatic hernia (CDH),15,21,24,25,28,29 chorioamnionitis,27,48 sepsis,33-36,39 gastroschisis,20,26 patent ductus arteriosus (PDA),19,22 persistent pulmonary hypertension,42 and respiratory distress.45

Table 2. Sample size for included studies using the SNAP-II

Illness severity as a predictor of mortality

Illness severity is closely tied with the concept of mortality risk. Clearly, as neonatal severity of illness increases, physiologic instability increases, and therefore the risk of neonatal death also increases. Not all of the included studies included analysis for this specific point, but much evidence is available to support that higher illness severity does lead to a higher mortality rate. Many of the included studies evaluated and reported the ability of the SNAP-II instrument to predict mortality in their study sample. Some researchers used the area under the receiver operating characteristic curve (AUC) to demonstrate the sensitivity and specificity of the SNAP-II. A result of 1 would be perfect discrimination, while a result of 0.5 would be completely random. The AUCs reported in these studies included 0.76,29, 0.77,15 0.81,41 0.82,39,47 and 0.88.12,40 Zupancic et al47 further divided the study sample into birth weight more than and less than 1500 g and found that the AUC was more accurate for the smaller babies (AUC 0.82 vs 0.79). Each of the results in the studies would be considered acceptable to good. It is interesting to note that the seminal article for the SNAP-II and SNAPPE-II instrument did not report an AUC for the SNAP-II, but reported only an AUC of 0.91 for the SNAPPE-II instrument across all birth weights.6

Some researchers used odds ratios (ORs) to describe how the risk of mortality increased along with increasing illness severity. Nakwan et al42 found that in a sample of neonates with persistent pulmonary hypertension, the odds of death increased by 1.04 for every 1-point increase in the SNAP-II score (OR = 1.04; 95% CI: 1.01-1.07; P < .01). Another study described the OR of mortality for newborns with respiratory distress (OR = 1.071; 95% CI: 1.040-1.103; P < .01). The third and final study described the increased risk of mortality in a cohort of neonates with CDH (OR = 1.057; 95% CI: 1.019-1.097).29 Risk of mortality was also described as a hazard ratio (HR = 1.09; 95% CI: 1.07-1.12; P < .0001)25 as well as the relative risk of mortality (RR = 1.07; 95% CI: 1.0-1.1).26 Nasr and Langer24 demonstrated that higher SNAP-II scores are associated with mortality (P = .005). Wilson et al21 demonstrated that illness severity scores (SNAP-II) predicted neonatal mortality (P < .001). Sundaram et al39 found that median SNAP-II scores were higher in the nonsurviving cohort (43 [36-53.5] vs 18 [16-37]; P < .001). Some studies used cutoff values to define mortality risk. For example, Dammann et al16 found that an SNAP-II score greater than 30 demonstrated that the newborn was 3.5 times more likely to die. Furthermore, when they adjusted for gestational age, the risk increased to 5.9.16

Illness severity and morbidity

Illness severity and morbidity are also closely related. Neonates who are sicker manifest increased illness severity scores and greater organ dysfunction.34,39 Neonates who are extremely low gestational age are also very much at risk for increased illness severity due to immaturity.16,17 When considering the relationship between illness severity and morbidity for neonates, researchers are also interested in the possibility of illness scores helping predict future morbidity, future treatment needs, and treatment outcomes.

Prediction of future morbidity

Critically ill newborns have common illnesses that occur perhaps because of immaturity or illness severity such as intraventricular hemorrhage (IVH) and necrotizing enterocolitis. Dammann et al17 found that SNAP-II scores greater than 30 were predictive of IVH and ventriculomegaly and echodense lesions in the brain; however, they did not see a relationship with cerebral palsy, autism, or microcephaly at 24 months corrected age of life. Chien et al31 found that the SNAP-II performed similarly to the original SNAP instrument and was able to predict IVH (AUC 0.73 +/- 0.02, [chi]² = 219) and chronic lung disease (AUC 0.78 +/- 0.01 [chi]² = 470).

Treatment needs

One study looked at severity of illness as a means to predict treatment needs for severely ill neonates. Specifically, Coleman et al15 found that 78% of CDH patients with SNAP-II scores greater than 25 required extracorporeal membrane oxygenation. Furthermore, SNAP-II scores were able to predict the use of extracorporeal membrane oxygenation (AUC 0.76; P = .003).

Treatment outcomes

A major area of focus is the relationship between illness severity and treatment outcomes. Wilson et al21 evaluated surgical treatment outcomes of CDH patients. He used the SNAP-II scores to compare the groups of patients offered surgical treatment and those who were not. Since more than 80% of the surgical cases with SNAP-II scores between 30 and 39 survived, and the potential surgical candidates had a 100% mortality rate, the authors were advocating for more specific guidelines of when to withhold surgical treatment from the newborn.21 Nasr and Langer24 worked to compare illness severity among CDH neonates who were inborn versus outborn, indicating whether the neonate was born in that facility or whether the newborn was born in another facility and later transported to the NICU. Although the inborn group was sicker than the outborn group (21 [IQR, 7-32] vs 5 [IQR, 9-12]; P = .0001]; outborn status was still associated with increased risk of mortality (OR = 2.8; P = .04).24 This seems to indicate that perhaps the quality of care in referral centers is not as optimal as within a tertiary center that routinely cares for the most critically ill infants. Capasso et al36 focused on outcomes of septic newborns who did and did not receive immunoglobulin along with antibiotic therapy. The SNAP-II was used to demonstrate that the groups were homogeneous for severity of illness prior to comparison of the outcomes with varying treatments. The weakness of this study is that they did not clearly report when SNAP-II scores were assessed.36 Madan et al19 sought to discover if severity of illness was related to successful PDA treatment; however, she discovered that treatment success or failure was related to gestational age. As gestational age increased, PDA closure rates increased (OR = 1.51 per week; 95% CI: 1.14-2.01; P = .004) and gastrointestinal complication rates decreased (OR = 2.41; 95% CI: 0.52-0.84).19

Illness severity as a control

As noted earlier, neonatal illness severity is a predictor of mortality and morbidity and because of this influence illness severity must be controlled when assessing outcomes between cohorts within and among NICUs. As the need for evidence-based practice continues, staff nurses will be called on to participate in more performance improvement projects as well as research studies. Accounting for and controlling for illness severity when evaluating outcomes in the NICU are a critical step to ensure accurate data interpretation. Eight of the included studies used the concept of illness severity, operationalized by the SNAP-II scores, as a means to establish group homogeneity or heterogeneity for illness severity before assessing outcomes between the groups. Included studies used illness severity as a control between septic newborns receiving antibiotics and immunoglobulin versus antibiotics alone,36 newborns receiving surgical ligation for a PDA at a cardiac versus noncardiac pediatric surgical center,22 newborns who had documented specific blood gas targets versus those who did not have specific blood gas targets established,25 neonates admitted to the NICU during the daytime versus nighttime,30 and finally 2 studies focused on inborn versus outborn status for neonates.24,32 One study used illness severity scores along with birth weight to match the control group to a group of neonates who were diagnosed with CDH. Since the purpose of the study was to compare costs between CDH and non-CDH patients, it was important to ensure that homogeneous samples were recruited for the comparison to be accurate.28 Finally, the extremely low-gestational-age newborn study used severity of illness, operationalized by SNAP-II scores, to adjust for NICUs that had a greater percentage of sicker newborns as this would skew the mortality rates for comparison.16

SNAP-II scores have also been used as a control when evaluating relationships between other variables. For example, Soraisham et al27 found a relationship between chorioamnionitis and early sepsis (OR = 1.6; 95% CI: 1.16-2.21) and IVH (OR = 1.6; 95% CI: 1.16-2.21). After controlling for illness severity, the relationships remained (sepsis: OR = 5.54; 95% CI: 2.87-10.69; and IVH: OR = 1.62; 95% CI: 1.17-2.24).

Illness severity on the first day of life

Most of the studies focused on assessing initial illness severity on the first day of life. Clinically this makes sense as one is establishing initial risk at birth. Specifically, this review seeks to establish the utility of the SNAP-II instrument, but it is important to note that many times when the researchers desire to report initial illness severity, they will report the SNAP-II with the perinatal extension (SNAPPE-II). This instrument accounts for 3 additional perinatal risk factors that will remain constant. That is why this is assessed only at 1 time point and not sequentially. Furthermore, most of the studies used a 12-hour data collection window for the SNAP-II score per the initial instrument design.6,16,18,20,25-32,41,42,45,47,48 However, 3 studies expanded the data collection period to 24 hours.15,23,43 Finally, 3 studies did not clearly report when the assessment was done and how many hours were used for the data collection window.21,22,36

Illness severity at later time points

Conceptually, illness severity is not only appropriate to evaluate on the first day of life. In fact, illness severity is something that will change over time. This was a primary reason why the author of the instrument separated the SNAP-II from the SNAPPE-II. The SNAP-II measures the physiologic state of the newborn and has the ability to capture the continuum of illness severity as it relates to physiologic instability. Therefore, it makes sense that a measurement instrument is needed that can quantify illness severity across the entire NICU length of stay. However, only 8 studies included severity of illness measurements after the first day of life. Two of those studies included more than 1 measurement; therefore, they will be discussed in the following section.

Two of the studies focused on illness severity assessment after a transport.38,40 Although the assessment of severity of illness could occur on the first day of life, oftentimes the assessment occurred later (gestational age at birth 35.3 vs gestational age at transport 35.7). Both studies also compared another instrument for severity of illness against the SNAP-II score. One study demonstrated that the SNAP-II, SNAPPE-II, and TRIPS (The Transport Risk Index of Physiologic Stability) performed equally well (P = .625).38 The goodness of fit was also tested across models and demonstrated a good fitting model using the HL goodness of fit (SNAP-II, P = .29; SNAPPE-II, P = .88; TRIPS, P = .49).38 The second study split the sample into surviving and nonsurviving neonates. The mean age of transport and subsequent admission to the NICU was 7.01 and 5.57 days, respectively, meaning that these assessments took place sometimes 5 and 7 days after birth. It is important to note that both instruments performed well. The AUC for the SNAP-II and TOPS (temperature, oxygenation, proxy for perfusion-capillary refill, and sugar-blood sugar) was .88 and .89, respectively.

The remaining 4 studies dealt with the assessment of illness severity with septic newborns. Sundaram et al39 sought to measure severity of illness for 12 hours after the onset of septicemia. The median age of these study participants was 4 days with an interquartile range of 3 to 6. Although the severity of illness assessment (SNAP-II) scores were delayed by 4 days for most babies, this study still demonstrated the ability to predict mortality (AUC 0.82; 95% CI: 0.68-0.95).39 This study also demonstrated that with SNAP-II scores more than 40, there was more organ dysfunction (3.8 +/- 0.4 vs 2.9 +/- 0.8; P = .001).39 This supports that the SNAP-II instrument was able to discriminate between varying levels of severity of illness. The final 3 articles sought to measure severity of illness at the highest level of clinical illness.33-35 The participants in the studies ranged in age between 1 and 15 days,35 within 2 days of sepsis diagnosis (range of diagnosis = 3-264 hours),34 and between 1 hour and 32 days.33 There is much diversity within the actual day of assessment for these studies. Not only is there variation among the 3 studies, but also there is variation within each study. The main purpose of these 3 studies was to assess relationships between SNAP-II scores and other biomarkers that could potentially provide additional information about illness severity.

Illness severity as a sequential measure

The main purpose of this article was to synthesize the available evidence measuring severity of illness prospectively. Of 35 studies, this review found only 2 studies that measured severity of illness over time. The first study measured the concept at 2 time points. Specifically, the author sought to determine whether there was a relationship with illness severity and outcomes associated with the treatment of PDA closure and indomethacin treatment.19 SNAP-II scores were measured at birth and then again with the first dose of indomethacin administration. No relationship was found between SNAP-II scores at either time point (birth and with first dose of indomethacin) related to failed closure of the PDA (P = .22; P = .82) or gastrointestinal complications (P = .24; P = .61).19 The relationships that they found indicated the critical influence that gestational age plays when considering treatment success and treatment complications for this population.

The final study that measured severity of illnyesess over several time points used initial severity of illness scores with the SNAP-II and SNAPPE-II for the first 12 hours of life and then continued to measure daily SNAP-II scores until the baby was discharged or died. The authors were specifically evaluating the relationship between severity of illness as measured by the SNAP-II and sepsis, necrotizing enterocolitis, and death. The study failed to demonstrate a relationship between SNAP-II scores as an accurate predictor of these later adverse events. In fact, results demonstrated that the majority (66%) of the SNAP-II scores greater than 10 were not related to any adverse events.18 And, the majority of patients who experienced a complication had a score of 0 within 5 days of the event. Furthermore, 92% of the total SNAP-II scores were 0; thus, the median SNAP-II score was 0.18

DISCUSSION

Clearly, the SNAP-II instrument has evidence to support its utility in neonatal research. Quantifying illness severity is helpful in research studies because it allows (1) quantification of the baseline risk of mortality and morbidity for the neonate, (2) quantification of how physiologically ill or unstable the newborn is at the time it is measured, (3) evaluation for homogeneous versus heterogeneous samples within and among research studies, (4) stratification of patients based on level of risk, and (5) the ability to determine initial risk at birth and then subsequent improvement or decline in the health of the neonate.

Neonatal severity of illness as operationalized by the SNAP-II has the majority of evidence available to support its sensitivity, specificity, and utility as a measure of illness severity measured during the first 12 hours of life or upon admission to the NICU. Although there is a recommended 12-hour data collection period for the SNAP-II,6 subsequent researchers have expanded this back to the original 24-hour window utilized by the original SNAP instrument.15,23,43 There is concern that treatment effects may be measured as well as initial illness severity with such a long data collection window. Although the data supported the validity of the SNAP-II instrument across all gestational ages and birth weights, Zupancic et al47 demonstrated that the SNAP-II is better at discriminating illness severity in very low-birth-weight neonates (<=1500 g).

The specific research interest of this review was to determine whether the SNAP-II instrument could be used to measure neonatal illness severity at a later time point (after the initial day of birth) and as a sequential measure. Two studies evaluated severity of illness after transport and provided evidence of validity and discrimination.38,40 The 4 remaining studies evaluated severity of illness in relationship to neonatal sepsis. Each of the studies found relationships that provided evidence of measurement of severity of illness. Sundaram et al39 found that SNAP-II scores were higher in babies who died and specifically SNAP-II scores greater than 40 were associated with more organ dysfunction.39 The other 3 studies were conducted by Figueras-Aloy et al33 and examined relationships between illness severity and various biomarkers associated with sepsis. Each of these studies demonstrated positive relationships to indicate that the SNAP-II does have some utility as an adequate measure after the first day of life. However, these were all single-center studies and involved only transported and septic newborns. There is potential that these newborns were sick enough to pass a threshold of discovery by the SNAP-II instrument.

Only 2 studies evaluated neonatal severity of illness over time in a prospective manner. Both Madan et al19 and Lim and Rozycki18 were not successful at finding hypothesized relationships. Only 1 study measured severity of illness daily and specifically sought to evaluate the ability of the SNAP-II instrument to measure illness severity over time.18 An interesting phenomenon surfaced concerning null values. Lim and Rozycki18 found that 92% of the SNAP-II scores were 0 and that null values often surrounded major morbidity events. One possible explanation is that if a specific test, for example, an arterial blood gas, has not been performed, then the assumed value is normal and therefore zero points are earned for the neonate. Treating missing data as normal values is questionable.

Strengths and weaknesses

A major strength of this review is that it included all of the research studies that utilized the SNAP-II instrument and was not limited to a specific age or date range. This is the first review article evaluating the utility of the SNAP-II instrument. However, the major limitation of this review is that it does not contain a review of all instruments that have been used to operationalize the concept of illness severity.

The strength of the evidence is supported by several large studies conducted across multiple hospitals. For many of the studies, this included neonatal networks that are working to promote research across the hospitals within specific geographic locations. The greatest benefit of this collaboration includes recruiting patients from different NICUs, so that the recruited sample might be more representative of the general population, thus more generalizable research findings. However, when recruiting patients across multiple hospitals, it is important to consider the effects that the environment may have on each subset of neonates. If the environmental effect is not controlled for, an ecologic fallacy error can be made when interpreting results and population-level findings are mistakenly attributed to individuals.49 As a minimum standard, the interclass correlations should be reported and provide justification as to why multilevel modeling approaches were not used. Perhaps the greatest weakness of the studies is that they were observational in nature with mostly convenience samples. This study type limits the ability of researchers to assess for cause-and-effect relationships among the study variables. However, the studies provide useful data within the ethical constraints of human subject research on vulnerable neonatal populations. Finally, although the studies provide measures of validity for the SNAP-II instruments, no reliability measures were reported.

Implications for research and practice

The SNAP-II instrument has been used successfully to quantify the concept of illness severity upon admission to the NICU; however, more evidence is needed concerning the sensitivity and specificity of the SNAP-II instrument when measuring illness severity sequentially and at later time points. Caution is advised when utilizing the SNAP-II for sequential measurements.18 Although there is evidence to support the validity of the SNAP-II instrument, reliability data are still lacking. Future studies using the SNAP-II instrument should report measures of reliability such as the Cronbach alpha for this purpose.

To date, the SNAP-II instrument has been used only to generate population-level data for research purposes.16 One author mentioned that illness severity scores can be used to counsel parents18; however, other authors have advised caution and recommended that SNAP-II scores not be used to guide decisions for individual patients.6,47 However, there is a need for further development of this tool so that it can provide individual patient-level data and it can be utilized as a clinical decision-making tool. Furthermore, researchers need to work with informatics specialists to automate the scoring process using data that are already routinely entered by neonatal nurses into an electronic medical record.47

CONCLUSION

Neonatal severity of illness is a key concept when caring for critically ill neonates. The SNAP-II has assisted researchers and clinicians in quantifying this concept to control varying severity of illness when evaluating outcomes at the population level. Although severity of illness is something that changes over time, more evidence is needed to determine if the SNAP-II instrument is able to accurately measure this at a later time point and as a sequential measure. Furthermore, the need for a precise measurement tool that can measure this concept at the individual patient level as well as the population level is the ultimate goal for improving neonatal critical care.

References