Background of the Invention Monitoring and treatment of premature infants or critically ill newborns is complicated and expensive. The efficace of the monitoring system, among other things, depends on which parameters are being monitored, and how many of the measurements can be made automatic without human intervention, without sacrificing system reliability. Several approches to measurement of illness severity have been known to be used hitherto, and those have varying degrés of cost and reliability. Most known approches have been devise, however, without specific emphasis on the patient to be monitored being a neonatal intensive care unit (NICU) patient.

Description of Prior Art Nearlv all IC (intensive care) illness severity scores based on physiologic derangements have been directlv or indirectly derived from the Acute Physiology and Chronic Health Evaluation (APACHE) score by Knaus et al. (see Knaus et al.

"APACHE--acutee physiologically based classification svstem"Crit Care Vled, 1981:9(8):51-597). He reasoned that derangements from physiologic norm are a measure of illness. and the more severe the deranaements, the more severe the

illness. He selected and weighted 34 vital signs and laboratory results routinely available in the first 24 hours of admission to form the APACHE. He showed that higher scores correlated with death, morbidity, and resource use (see Knaus et al.

"APACHE--acutee physiologically based classification system"Crit Care Med, 1981;9(8):591-597). The advantage generally of such physiology-based mesures is that they are objective, reliable. and credible. APACHE nvas simplifie to APACHE-II (see Knaus et al. "APACHE II: A severity of disease classification system" Crit Care me, 1985; 13 (10): 818-829), and was copied and simplifie into the Simplifie Acute Physiology Score (SAPS) (see Le Gall et al."A simplifie acute physiology score for ICU patients"Crit Care. Med, 1984; 12: 975-7). With increasing sophistication, each of these was revised into APACHE-III (see Knaus et al."The APACHE III pronostic system. Risk prediction of hospital mortality for critically ill hospitalized adults"Chest 1991 ; 100: 1619-36) and SAPS-II respectively (see Le Gall et al. "A new simplifie Acute Physiology Score {SAPS II} based on a European/Ntorth American multicenter study"J_AlyIA 1993; 270: 2957-63).

In pediatric intensive care, APACHE was modifie to create the Physiologic Stability Index (PSI) (see Yeh et al."Validation of a physiologic stability index for use in critically ill infants and children''Pediatr Res 1984; 18: 445-45 1), which was then simplifie to create the Pediatric Risk of Mortality Score (PRISE) (see Pollack et al."Pediatric Risk of Mortality {PRISM} score"Crit Care Med, 1988; 16: 1110- 1116), and later the PRISSvl-III (see Pollack et al. "PRISM III: an updated Pediatric Risk of Mortality score"Crit Care Med, 1996; 24 : 743-52) and PRISM-III APS (for "Acute Physiology Score") (see Pollack et al."The Pediatric Risk of Mortality III-- Acute Physiology Score {PRISM III-APS}: a method of assessing physiologic instability for pediatric intensive care unit patients'J Pediatr 19971 131 :7-81).

Proir Art: Illness forNewbornsScores The importance of measuring illness severity became clear in neonatal intensive care which prompte similar score development. In 1993. Richardson, et al. used the APACHE concepts but all new variables in creating and validating the Score for Neonatal Acute Physiology (SNAP) (see Richardson et al. "Score for Neonatal Acute Physiology {SNAP}: Validation of a new physiology-based severity of illness 1993;91:617-623).WilliamTarnow-Mordi.etal.usedtheindex"Pedia trics

PRISM concept and all new variables in creating and validating the Clinical Risk Index for Babies (CRIB) in 1993 (see International Neonatal Network "The CRIB {clinical risk index for babies scores-a tool for assessing initial neonatal risks and comparing performance of neonatal intensive care units"Lancet 1993 : 342: 193- 198). SNAP is a 34-item physiology-based score measuring severity of illness, applicable to all newborn intensive care unit (NICU) admissions.

Based on the same concepts published for APACHE and PRISM, Dr. William Tarnow-Mordi derived and validated the CRIB (Lancet 1993: 342; 193-198) for very low birth weight infants (<1500 grams) treated in NICUs in Great Britain. CRIB uses only three physiologic variables, derived from routine vital signes and laboratory values, along with three standard markers for newborn risk, i. e., birth weight, gestational age and the presence of a lift-threatening congenital anomaly. CRIB is in widespread use for research in Europe. There are, however, several important CRIBshortcomingsof a. Validated onlv for verv premature infants (<1500 m) : While this is valable for research purposes, it is unacceptably restrictive for hospital and ICU managers who need to assess performance of all admissions, not just a special subset. The present invention, in contrast, is validated for all birth weights. b. Ouestionable performance for outborn babies: The CRIB makes assumptions about incomplete records that are untenable for infants under- emergency transport conditions. The present invention begins scoring only after the infant enters the ICL ;. thereby avoiding measurement assumptions. c. Admission-onlv score. no seauential scoring: CRIB is designed to reflect severity onlv in the first day of life. Half of its components are fixed at birth (birth weight, gestational age, presence of anomalies). The present invention has a much broader dvnamic range, designed to measure changing condition over time. All of the adult and pediatric scoring systems have this broader dvmamic ranae characteristic.

d. CRIB'sphysiologicitemssampleonlytheorgan-svstem: respiratory system. This may be adequate in a homogeneous population of very premature infants all of whom will have degrés of respiratory failure as their illness on admission. For full term infants, a much wider variety of organ-system failures requires a broader sample of organ-system items. The present invention samples several organ systems.

The"Berlin"score is a recently reporte German score (Maier RF, Arch Dis Child 1997; 76: F146-F151) and is more of an epidemiologic adjustment tool than a true neonatal illness severity score. It too is an admission-only score and applies only to very premature babies (<1500 grams).

There are several other important adult ICU scores that require mention, because they have used parallel techniques to construct the scoring systems.

Mortalitv Prediction Model (MPM): A series of mathematically sophisticated adult ICU risk order was developed by Teres and Lemeshow (Care Med 1987; 15: 208-213). A brief attempt at commercializing these appeared to be unsuccessful.

S., KPS I and SAPS II: The unwieldiness of the original APACHE led to an independent revision by LeGall et al. (Le Gall JR: Con't Care Med 1984; 12: 975-7) into the Simplifie Acute Physiology Score (SAPS), creating a direct competitor to the concurrently derived APACHE II. Later, with the collaboration of the inventors of the MPM, the SAPS inventors revised and simplifie their score to create SAPS II (LeGall JR: JANIA 1993; 270: 2957- 63). The S.KPS has been developed in France and used widely throughout Europe.

A need still exists in healthcare to provide a system and method which obviates the disadvantagesofknownsystemsandmethodologyforshortcomingsand determinina seerity of illness of patients, in particular, neonatal ICU patients from the moment of admission.

Summarv of the Invention It is an object of the invention to provide a computerized method and svstem for determining severity of illness of a neonatal patient, using a computer method and process, and other measurement hardware.

The invention, in its broad form, resides in a computerized method of making an on- line determination of illness-severity of a neonatal patient in a predetermined time span, bu usinez a software proaram and optimal weighted measurement values of a predetermined 'n' number of on-line parameters from the patient beina monitored, said parameters relating to n mesurable predetermined physical conditions, said method comprising the steps of: (a) obtaining, in said predetermined time span from the neonatal patient being monitored, several carefully selected values of each of said n physical conditions and producing, using a proffram, a single optimal value from said plurality of measured values for each of n measured physical conditions; (b) usina said softtvare program, obtaining from said single optimal value a modifie weighting partial score, thus generating n modifie weighted partial scores for n on-line parameters from the patient being monitored; and (c) summing the n modifie weighted partial scores to provide an indication of illness-severitv of the neonatal patient, which severity indication can be displayed and compare with other known values from databases.

It is another object of the invention to provide a computerized method and system for measuring a mortality risk level of a neonatal ICU patient from the measure of illness severity.

Applicants have reviewed the measurement of illness severity in nelvborns and concluded that no comparable scale development has occurred in neonatal intensive care until now. To a larae deoree, birth wellht has successfully served as a proxy indicator for severitv of illness alona with other risk factors including gestational age. sex and race. Nonetheless there is strong evidence as presented above that

these prior art models fail tO accurately capture severity of illness as evidenced by the larae residual variation in studies of mortality and newborn lung disease.

Used in applicants'invention is a selection of scale items, and scoring of therapies and physiologic deranaements. Applicants followed the APACHE convention of mostseverelyderangedphysiologyina24-hourperiodfollowingabstr actingthe admission. The hypothesis to be tested was that the traditional risk factors (including birth weight, Apgars, sex, race, etc.) were strong predictors of mortality across the birth weight spectrum, but inadequate for distinguishing mortality risks within birth weight strata.

A comparison of variations in outcomes and resources is one application/use of the present invention. Another use is the refinement and improvement of the prior study. In as much as SNAP-II is a scoring system to classify illness severity in newborns in intensive care units (NUCUs), it mesures the degree of physiologic derangement across multiple organ systems, usine vital siens and laboratory values and other information routinely recorde in clinical records. Five core components of the present system, which are significant, are: l. Score items: Of the hundreds of potential markers of newborn illness severity, Applicants have selected a very limited subset that are reliably available, easily capture, and robust as predictors. This specific list of variables, and their exact definitions, are unique to the present invention improvement over SNAP.

2. Items score weights : Each score item is weighted according to a carefully derived risk value, so that less serious items have relativelv low score points, and more serious items have relatively high score points. The SNAP-II is the sum of points for each item. These variables and weights have been derived on an initial large cohort of patients, and then validated on a second large cohort. The present invention improvement takes into account t VO subscores.

oa. Risk factors: These are standard, scientifically recognized risks for neonatal mortality, including birth weight, gestational age, Apgar scores and amender. b. Acute Phvsiology Score: These items reflect the seventy of physiologicderangement. équations and coefficients: These equations associate the SNAP-II (or its individual components) with an array of outcomes, resource use, costs and process benchmarks. They are derived from régression equations (liner, logistic, polynomial) through a model fitting process that involves modification of the input variables and selective inclusion to optimize the discrimination and calibration of the equations.

4. Reference database: This database consists initially of the combine NICU data sets from the 7 NICU research study in New Enalarid, funded by AHCPR, and the six NICU study in California, funded by Kaiser Permanente Division of Research. and the 18 INICU data set of the Canadian NICU network, founded by the Canadian Nledical research council.

5. Reportino sn stems: These reports lay out risk-adjusted comparisons of pre-specified groups of patients in an individual WICU with an appropriate reference group, drawn from the reference database.

The present invention improvement is unique compare to other newborn severity scores in several ways: it is shorter and simpler to use than the original SS=AP. It applies to all newborns admitted to NICUs, in contrast to the CRIB score which applies onlv to babies weighing <1500 grams. It is adaptable to sequential scoring, so that nez scores can be generated daily. In contrast, CRIB is an admission-only score. The present invention is distinct from the adult and pediatric illness severity scores of prior art in that it applies specificallv to newborns in ìsiICUs. The APACHE I, II, and III, and SA-PS I and Il are not applicable to newboms. The PRISM, PRISM Il. and PRISM III APS were derived using some full term

newborns but their applicability and calibration specifically for premature infants has never been evaluated.

Thus, even though in prior art. several attempts have been made to provide svstems which are intended to provide an indication of the illness severity of a patient. there have been the following sianificant disadvantaaes in applying them to neonatal care situations: -in some prior art systems, as many as 34 different physical condition measurements were used, thus making the process very elaborate, expensive, and prone to miscalculation; -efforts to reduce the number 34 were mande, but an identification of those physical conditionsiparameters which would be crucial to neonatal situations was not made based on available illness severity scores from similar patient population ; -A concerte effort was apparently not made in prior art to make an automate selection and optimization from several available readings or measured values of a single physical condition for neonatal applications over a predetermined time span.

The choice of the single optimized selected measured value of a given physical conditiory parameter in the present invention is so made bv a program that the choice is linked with the most appropriate weighting factor based on vast amounts of accumulated prior data, and the choice additionally ensures the highest possible reliability of the severity measurements generated. Certain physical conditions selected for being monitored in prior art were not ideally the best suitable physical conditions for neonatal measurements.

For instance, in the present invention, in a preferred embodiment, the physical conditions chosen for monitoring are blood pressure, temperature, oxygenation ratio. serum pH presence, absence of seizures and urine output. The program in one preferred embodiment makes the optimal choice of those parameters as the lowest mean blood pressure, the lowest temperature, lowest oxygenation ratio, the lowest serum pH, presence'absence of seizures, and urine output. Consequently, in at least three of the six phvsical conditions selected for neonatal monitoring, four of the

physical conditions offer optimal selection by the proaram resultina in the choice of the lowest of measurements. The fact that the lowest of the measurements for blood pressure, temperature, oxyaenation ratio and serum pH were selected by the program is also associated with certain predetermined weighting factors which would result in the most reliable illness severity determination when compared with available data of similar patient population.

Alternatively, if instead of the lowest measured values of blood pressure, temperature, oxygenation ratio, and serum pH, other values were to be chosen by the program, the weighted factors to be used would correspondingly be different, in order to ensure reliability of the resulting illness severity measurement.

Applicants have found the use of the lowest mean blood pressure, lowest serum pH, lowest temperature, lowest oxyQen (pAO/FiO) ratio, as well as urine output and the presence of multiple seizures, along with the weighted factors disclosed herein would provide an extremely hijhly reliable measurement of the illness severity for neonatal patients.

The task of selectina a single optimal measurement value from several generated measurement values of a physical condition, and the task of matching up the selected optimal measurement based on the nature of the optimization, for e. g., the selection shaving to be the lowest value with an appropriate weighting factor, is done by a computer program in the present invention.

Brief Description of the Drawinos The foregoing and other objects, features and advantages of the invention will be apparent from the following more particular description of preferred embodiments of the invention, as illustrated in the accompanying drawings in which like reference characters refer to the same parts throughout the different views. The drainas are not necessarilv to scale, emphasis instead being placed upon illustrating the principles of the invention.

Figure l is a schematic system diagram of a preferred embodiment of the invention illustrating neonatal patients beinL7 monitored and with a vital

signs interface. programmed computer, database, and a laboratory interactino, Figure 2 illustrates the process flow of how data is acquired on a neonatal patient admitted to the WCU is monitored usine the principles of the invention, Figure 3 illustrates how from several different changing values of a given physical condition/parameter, an optimal measured value can be selected using the program in the invention, Figure 4 illustrates an alternative embodiment for the inventive system illustrating a processor including the monitoring software and its interaction with inputs and outputs as shown, Figure 5 shows a scatterplot of SNAP-I versus birth weight, Figure 6 shows a scatterplot of SNAP-II versus birth weight for comparison with 5, Figure 7 is a bar-graph of mortality by birth weight and illness serve' Figures 8 shows mortality by NAPPE-ICI score, Figure 9 shows a correlation between SNAP-I and SNAP-II scores.

Detailed Description of Preferred Embodiments Figure 1 illustrates components of the svstem of an embodiment of the invention, wherein neonatal patients admitted to the fICU 101 are monitored to obtain measurements of standard vital signs and/or other predetermined parameters such as. for example, blood pressure, temperature measurements, urine output, and presence or absence of multiple seizures. These individual values are collecte 102 and registered on the computerized (or paper) patient records. Laboratory results on individual patients are similarly registered on patient records.

At fixed time intervals, all readings of the six key physiological parameters are obtained and transmitted to the computational software 105 as necessary. A measured parameter may have several recorde values in the specified time interval.

The worst value, for example, is selected using careh. illy defined predetermined criteria. The degré of derangement is ascertained and a partial score is assio-ned for each parameter. In the illustrated embodiment, the weighting consists of multiplying each worst parameter with a predetermined score for the measured parameter.

The program 105 advantageously chooses one of different methods of optimization such as exemplified in Fics. 3. Depending on which method of optimizing is chosen, for example, depending on whether the maximum measured parameter value in a given time span is choses or the minimum 303 lowest logical value or the weighted averaae 302 is chosen, a corresponding weighting factor for multiplication is lined up by the program 105 for multiplication. Advantageously, the program 105 is made to ignore or trap a measured value which is illogical. By user-intervention, command default and the like, program 105 is made to proceed with whatever kind of optimizing is desired. The kind of optimization so chosen will have its corresponding predetermined multiplication weighting factors based on which zone the measured physical condition values lie in.

In the preferred embodiment described herein, as stated earlier, the parameters chosen are the lowest mean blood pressure, the lowest pH, the lowest temperature, the lowest oxygenation ratio, urine output, and presence of multiple seizures.

It is conceivable that for optimal values, aérage values of measured predetermined parameters (physical conditions) may be used in the system and method of the invention if so chosen by the user. In such an event, weighting factors stored in the database 106 corresponding to aérage values of the predetermined measured physical conditions will be used. Advantageously, the weighting factors may automatically be verified as to their applicability before use in the illness severity measurement.

In an extension of the preferred embodiment referenced above. three additional physical conditions (factors) are measured and considered after weighting as

controlled by the software in the program 105 of Figure 1, to assess the mortality risk of an NICU patient being monitored. The three-factor addition is termed as Perinatal Extension. or PE, which, when added to SNAP II, will provide SNAPPE II.

Described in the following passages are some considerations which governed the choice of physical conditions and their optimal selection.

The original SNAP was cumbersome and time-consumina, limiting its use predominately to research. With the example of the successful simplification of APACHE, SAPS, and PRISE, applicants have revised and simplifie SNAP into SNAP-II. Applicants had to bear in mind that the spectrum of illness and causes of death and other considerations are different in neonatal patients as compare with pediatric/adult situations. Accordingly, for NICU patients, the preferred embodiment of SNAP-II described herein is a six item physiology-based severity of illness measurement that has predictive performance that equals SNAP. Consistent with"parent"physiology scores, it uses readily available vital suons and laboratory mesures obtained from routine medical records. It provides an objective, reliable measure of the initial status of an NICU patient at the time of NICU admission.

SNAP-II advantageously consists of six related components.

I. Item selection: The 6 items contained in SNAP-II are listed in table l. These were carefully selected to include devinable, mesurable items that maximize discrimination between sick and well newborns. Applicants, after research, have retained electronically accessible items and eliminated unreliable, weak, or infrequent items.

TABLE 1 Items Included in the SNAP-II Scores (SNAP-II)Physiologicalitems lowest mean blood pressure lowest pH lowest temperature lowest pAO,/FiO2 ratio urine output multiple seizures Peu'natal extension SNAPPE-II birth weight small for gestational age (3rd%) low Apgar score Mortality risk computed from aboie

2. Specific definitions: These definitions and abstracting rules are clear and simple, ensuring accurate data acquisition. Data are collecte either prospectively or retrospectively from the medical charts corresponding to the first 12 hours of admission. starting at the time of the first set of vital signs on admission and extending for exactly 12 hours. The single worst value for each item is selected and recorded. ensuring consistent reliability amont abstractors. If a test were not ordered, it is assumed the clinician estimated it near normal and not needed to direct care. These items and other measured items, which are near normal, receive zero points.

3. Scorina ranaes of each item: A gradient in risk for each item is reflected by derangementsintozones(seebelow).Thesezonesarecontiguousdivid ingthe ranges and have been optimized in their power to discriminate between survivors and non-survivors. For convenience, applicants have designed a data form to simplifycollection.

4. Point score for each item-ranCe combination: Points are then awarded for each item according to degree of derangement. These points are optimized to predict mortalityl"sur-vival.

TABLE2 SNAP-II'i Score (Preferred Embodiment) Parameter Points ModerateSevereExtremePoints >30 20 - 29 <20 BloodPressureLowestMean 0 9 19 LowestTemperature (F°) >96 95-96 l5 o 8 15 -2.49.03-.99<.03>2.51.0 pAO2/FiO2 Ratio 162805 -7.19<7.1#7.207.10 LowestSerum pH 0 7 16 None/SingleMultiple Seizures 019 >0.9>0.90.1 - 0.9 <0.1 Urine Output (cc/k/hr) 0 5 18

5. SWAP-IN: The sum of points awarded for each item quantifies the degré of physiologic derangement and creates a unique marker for overall illness severity.

This measurement is the SNAP-II illness severity measurement.

SNAP-11 = sum of points awarded for the following: (low mean BP) + (low temp) + (low oxygenation ratio) + (low pH) + (multiple seizures) + (low urine output) 6. SNAPPE-II : SNAP-II mesures illness severity, a critical component of mortality risk. However, other factors are known to make the risk even higher.

Applicants tested numerus risk factors following the same process as 1-6 above, and selected three additional perinatal items (sexe table below), developed specific definitions and score ranges, and assigned supplemental points. The sum of SWAP- II plus the Perinatal Extension points creates the SNAPPE-II, an estimate of mortality risk.

TABLE 3<BR> <BR> SNAPPE-IITi Score NoPointsParameterPoints Moderate Severe -900<#100gm750 Birth Weight (gm) 0 10 17 <3rd#3rd Small for Gestational Age 0 8 10<77- Apgar at 018minutes

NAPPE-ICI = sum of points awarded for the following: SNAP-II + (BW points) + (SGA points) + (low Apgar points) 7. SNAPPE-II mortality risk : The points for each item of SNAPPE-II represent the coefficients from the logistic regression equation predicting mortality. This process can be inverted by scaling the coefficients down by a factor of 0.783, subtracting the constant, and exponentiating the mathematical term. The result is a point estimate of individual mortality risk, P (death): P (death) = 1/(1 + eR) where R = (SNAPPE-II * 0.783)-5.1656 SNAPPE-II is defined as: 9. if moderately abnormal low mean blood pressure OR 19, if severely anormal low mean blood pressure + 8, if moderately anormal low temperature OR 15, if severely anormal low temperature + 5, if moderatelv anormal oxyoenation ratio OR 16, if severely anormal oxvgenation ratio OR 28, if extremelv abnormal oxvenation ratio + , if moderately anormal low pH OR 16, if severely abnorrnal low pH + 19. if presence of multiple sizures +

5, if moderately anormal urine output OR 18, if severely anormal urine output- moderatelylowbirthweightingrams+10,if OR 17, if severely low birth weiQht in c-trams 8, if moderate smallness for aestational aQe + 18, if severe "Apgar" at 5 minutes.

Figure 2 is a flow diagram of an exemplary method of illness severity measurement according to the invention, and is self-explanatory. The embodiment in the Figure 9 flow dlaaram is not limited to six or nine parameters. At step 201, a neonatal patient to be monitored is admitted into the neonatal intensive care unit; the patient, at step 202 is connecte to the patient vital sign monitors, these in turn and to the computer and the database. At this time. arrangements are also made to obtain any manually entered physical condition readings which are entered into the patient medical records. It is within the scope of this invention to use any physical condition measurements, which can be made automatically, without any human intervention.

At step 20'), the n different physical items to be monitored are identified. This can be done manually, or can be done automatically by default or other methods of maki2 a choie. e. g., bv pushing a single button which is connecte to make a selection of'n', predetermined physical conditions to be measured. In block 204, all measurements of values for each chosen physical condition are extracted and fed into a buffer. Values determined to be illogical are then excluded. Then, for each parameter, the"worst"of the values is selected. Step 205 pertains to identifying the physiologic range that contains the value selected in 204. Even though in the preferred embodiment, three ; four zones are indicated for practicality/simplicity, alternatively, there may be two zones or five or more zones. Step 206 pertains to attributing weighing factors for each zone of a physical condition ; parameter. The software proram, after takino into account several cases for similar patient population as a patient in question. decides suitable weighting factors which are used as multipliers for an optimized value of the measured physical condition. Generally, the morse the phvslcal condition. the higher is the weiahtinQ factor. Typical weighting factors are shown in connection with a preferred embodiment in Table 2

and Table 3. In steps 207 and 208, weighted partial score for each measured parameter is chosen with the help of the software program. An example is illustrated in figure 3 where varying values of a measured parameter with respect to time are shown. Out of the available values, as examples, a maximum or minimum or an aérage or a root mean square value can be chosen by the sofîware program.

Other methods of optimizing are conceivable and within the purview of this invention. In step 209, the optimizing value is modifie to generate a modifie partial score for a measured parameter. In the preferred example, the modification consists in automatically multiplying the optimized value of a score with its associated weighting factor. Other methods of modifying are within the purview of this invention. An optimized value of a parameter may, for example, be divided by a suitable weighting factor to obtain the modifie partial score for the physical condition (parameter) in question. In step 210, there is an automatic addition performed of all the modifie partial computational values or scores to determine illness severity, Step 211 compares the determined illness severity with other comparable information for similar patient population. The data of illness severity measurements can be used as in step 212 to review and update future illness severity and mortality risk measurements and for other suitable purposes. Some of the possible implementations of the SNAP-II system are presented in this text under the subheading"Specific applications of SNAP-II". Other uses and applications not specificallv stated herein are conceivable and are intelligible to those skilled in the art.

Figure 4 illustrates an alternative embodiment of the inventive system showing the interaction of a processor, monitoring softnvare, and NICU patient measurement inputs.

Shaving collecte background data and generated working modules (the coefficients for supporting the models), Applicants formed the scorinQ system,apparatus for hospital neonatal ICI is and physicians. As illustrated in Figure 4, this embodiment of the present invention inclues three components--(i) monitorino sofavare, (ii) optional input accessorv and (iii) sytem environment.

The monitoring software in the figure 4 embodiment is at the heart of the invention (with its beginning in SNAP-II). The software generally is executable code or a circuit or combination thereof within a digital processor/computer processor 401 (represented by dashed line box in Figure 4). The software may include a database configured through known database techniques (i. e., prolrams such as Access, Parados, and the like). A user interface enables user input of SNAP-II data items which are then stored in the database according to the database program employed.

A processing member of the software is responsive to the data held in the database and calculates the SNAP-II score according to the working models of the present invention (and formulated by Applicants in the study described above). The software provides the SNAP-II score on output to the user through output means (e. g., a monitor 406, printer 409, etc.) Optionally, various front-end, input accessories 403,404 may be employed. This inclues hardware, software or combinations thereof that provide data to the database. For example, an optical scanner may be used to convert data in hard copy or paper form to electronic/digital form and store the latter in the database. If the database is coupled to a network or a aencral hospital system, a multiplicity of input streams may be involved. Thus, a cachiez subsystem and/or other clearing house subprogram may be employed to extract from the input streams pertinent SNAP-II data and store the same in the database 406.

The real utility will accrue to users 405 who harness the SWAP IN (invention) measurements with the predictive equations (working models) to produce variance reports on clinical performance, resource use and inter-institutional benchmarking.

There are txvo strategies for developing such NICU management information systems: a. Stand alone svstems. These are expansions on the scoring software described above to include equation-based benchmarks that generate regular reports for NICU managers. These systems may be quite comprehensive, and can be designed to accept data-feeds from hospital Admission ; Discharge. Transfer (ADT) evenhospitallaboratorysystems.suystms.or

b. Inteorated svstems: In the rapidlv consolidating market for electronic medical records, an alternative strategy is to partner with a major vendor of hospital electronic records. With their expertise and funin2, one can work to embed the SNAP II score into the electronic chart, providing automate data acquisition and scorina, and use this to prompt actions for the clinicians in real-time. Many of the decision-support processes would depend on SN XP and related data. but also derive from the wellorganizedcarepractices.quality, NIore specifically, there are several major vendors of integrated medical records systems: i. Hewlett Packard. CareVue Svstem: CareVue is an electronic charting system for intensive care units, including nursing flow sheets, physician notes and clinical pathways. Such a charting system is an ideal vehicle for an integrated system, because the electronic accessibility of all SNAP II components reduces data collection costs to near zero. In this setting, one can concentrate on developing the severity-adjusted variance reports and decision-support feedback in real time. ii. Cerner: Cerner is a major vendor of hospital laboratory data management and reporting software. It has expanded into a comprehensive data repository for clinical records. The system is state-of-the-art, flexible, based on networked workstations, and using web-based technology. The Cerner system has focused its development efforts on outpatient records, so it has no specific ICU charting software similar to CareVue. iii. Pices/Phamus : This is another major vendor of hospital comprehensive software. Others are suitable.

Figure shows a scatterplot of SNAP-I versus birth weight. Dark spots (diamonds) represent thatregardlessofthebirthweight,higherSNAPscoresshowing are associated with the patient death. Also seen is a concentration of dark spots at low birth weights regardless of the SNAP scores.

Figure 6 shows a scatterplot of SSAP-II versus birth weight. Darker circles represent death. It is seen that SNAP-II effectively discriminates between survivors and deaths across all birth wights.

Figure 7 is a bar-graph of mortality by birth wcl (zht and illness severity. confirmina findings from Fig. 5. Mortality rates for three different ranges of SWAP scores are illustrated. The graph shows the hiahest mortality rate for birth weiaht group of under 750 trams, even for SNAP scores of 20+.

Figures 8 shows mortality by SNAPPE-II score, showing gradient in mortality risk for increasina SN-APPE-II score values. This relates to data unpublished at the time of this application and is limited to the New England area of the United States.

Figure 9 generally shows correlation beriveen SNAP-I and SNAP-II measurements.

It is generally seen that patients with higher SNAP-II scores (thin circles) died. An excellent correlation between the SNAP and the SNAP-II scores is seen from Fic,. 9.

Substantative differences between the SNAP-IIF CRIB. SNAP. APACHE-III. and PRISM-III: This section highlights the differences amont the scores that may be relevant to the present invention. In summarv. CRIB and prior art SNAP are the most closely relate, but are public domain scores, and the latter was invented bv the applicants.

APACHE-III is proprietary, but apparently applies only to adult ICUs. PRISI-III relates to children, but serves a distinctly different population in pediatric ICUs rather than NICUs, and has been developed and validated for this different purpose.

1. CRIB: CRIB was developed and validated for premature infants weighing <1500 grams at birth andior delivered at <32 weeks gestation. It does not apply to larguer or older babies and therefore cannot be used to characterize the entire population of aNICU. The score is computed from the time of birth (not admission) which makes it difficult or impossible to apply to outborn infants (infants transferred to the current hospital). Two of the physioloaic items (highest and lowest "appropnate"concentration of oxygen administered) are extremely complicated to understand and to abstract and are therefore error-prone. Furthermore, these items

ocannot be obtained from electronic sources (i. e., laboratory computers). CRIB was validated on a cohort of infants in 1990 and has not been re-calibrated since, despite a well Recognized 50 percent fall in the mortalitv rates (see Richardson et al.

"Declining Severity Adjusted Mortality : Evidence of ImprovinQ NICU care'' Pediatrics. ; ? 998; 102:893-899}). These six CRIB items blur the distinction between the factors(birthweight,gestationalage.congenitalrisk anomalies) that do not change from day to day, and degré of illness (physiologic derangements including worst base deficit, and highest and lowest "appropriate" oxygen requirement). CRIB was openly published and is now in the public domain.

2. SNAP from prior art: SNAP is a 34-item score that is cumbersome and time- consuming. SNAP-II has eliminated 28 unnecessary or redundant items. SSTAP had a number of variables that are difficult and unreliable. SNAP used a 24-hour baseline data collection and scoring period. (The SVAP-II scoring period has been reduced to 12 hours to minimize"contamination"of scoring with responsiveness to treatment). SNAP weighted items based on estimates of an expert panel. (SNAP-II weights item empirically based on a logistic regression model). SNAP has supplemental points to recognize mortality risk not capture by physiologic derangement-the Perinatal Extension points of the SNAP-PE. (SVAP-II uses the same three items but these are now empirically wighted according to the logistic regession model). SNAP was openly published and is now in the public domain.

3. APPACHE-III : APACHE and APACHE-II were both openly published public domain scores. APACHE-III was developed as a private venture. The scores and definitions have been released publicly. The rest has been held as trade secrets.

This inclues the score ranges and points, the equations relating APACHE-III to a variety of outcomes, the computer programs developed to compute and display results, and the APACHE reference database. APACHE scoring applies only to adult ICTUS and is not applicable to children or particularly to newborns because of their verv different phvsioloay and diseases.

A. PRISSE: The PSI and PRISAI were openly published scores. and are in public domain. PRISNI-III is, as aforesaid. disclosed in published European application EP

0 764 914. 42. At least some of the distinction between SNAP-II and PRISM-III which are cn'tical to an understanding of the present invention follow: a. Patient population: SNAP-II claims applicability to all patients in newborn intensive care units (NICUs). PRISM-III claims applicability to all patients in pediatric intensive care units (PICOTS) which may include some newborns.

Patient assignment to NICU or PICU may vary slightly amont hospitals, but is dominated by age. PICUs serve predominantly older children and virtually no premature infants, whereas NICUs have all newborns, predominantly premature infants. Another critical difference in population between NICUs and PICUs is the range of diagnoses. PICUs serve a wide variety of conditions including infections, cancer, surgery, etc., whereas NICUs server predominantly premature infants, or term infants with disorders of fetal-neonatal transition. congenital anomalies, and birth accidents. b. Score items: Each score has selected and optimized items to predict mortality in the respective ICU setting. The definitions and score ranges reflect the physiology of the respective age groups and technologies and practice patterns in . pets of ICUs. PRISM-III adjusts for diagnoses; and the current SNAP-II does not. c. Scores: The scores for the different methods of illness severity measurement in prior art are the sum of the individual selected parameters and are therefore not related nor compatible with illness seversity measurements from the present invention. d. Eauations: W hile the forms of the equations are similar (both derived from the logistic models) the coefficients and constants are different and optimized to the populations and diseases of the respective ICTUS. e. Reference databases: Applicants believe that there is nothing proprietary regarding databases for either product : both retain control through limited access and trade secret protections. In both svstems, the reference database is vital for providing benchmarking of performance on several outcomes.

Advantaaes of the Present Invention 1. Priority in newborn scoring : Applicants believe they were the first to develop and report physiology-based illness severity scoring for newborns with SNAP.

SNA-P-II is derived from SSt. XP.

2. Unique items selection: Of the hundreds of potential predictors, applicants have selected the 6 that best capture illness severity, and include items from each organ group.

3. Unique score we* ahts: The score weights are based on the logistic regression model used to derive the score. These weights represent optimized mesures of mortality risk.

4. Unique scores for SiNAP-Il and SNAPPE-II: The summary scores are computer determinations from unique components to create a unique combination, reflectina multi-system physiologic arrangements and mortality risk.

5. Unique mortality risk equation. This equation was derived and validated on over 27.000 cases, ensuring a precise and robust estimate of mortality risk.

6. Unique computer proaram. The items, definitions, scores and computations are embodied in a first aeneration computer program that calculates scores and mortality risks for populations of patients in newborn intensive care.

Implementation of the SNAP-II svstem SNAP-II. SNAPPE-II and SWAP-IN mortality risk are intended as an integral part of a comprehensive WICU outcome evaluation and reporting system. Applicants are committing each component to a software implementation. both as stand-alone programs and embedded in larder commercial ICU charting and decision support proarams. The components of applicants'system are: 1. NICU admissions and outcomes database. It is essential to record the key risk-factors. treatment processes and outcomes of all NICU admissions. An example of this is the Neonatal Minimum Database Svstem (IjMIDS) which contains a strictly

olimited set of the key risk factors, e@ents and outcomes that permit full characterization of the performance of a given NICU.

2. SNAP-II. The collection of the 6 items ? plus the Perinatal Exlension risk factors permit measurement of severity of illness and mortality risk on admission.

3. Risk adjustment of outcomes. The risk-adjusted performance of a NICU can be calculated for each of several outcomes. Each risk adjustment equation is intended for inclusion in this patent application. The process of risk adjustment is carried out as follows. Individual risk factors for each patient are combine with the SNAP-II on that patient to generate an individualized risk (i. e., a probability between 0 and 1) for that specific outcome. The sum of these risks for a designated population represents be"expected"incidence of that adverse outcome. The actual or"observed"incidence for that same population is then tallied and compare to the expected rate. The ratio of the observe to the expected rate is called a standardized rate, and is used for comparisons of performance (see below). Applicants have developed risk-adjustment equations for the following: a. Mortality. Mortality was used as the standard for developing the SNAP-II score. It is possible to calculate standardized mortality rates for any size population (see equations above). b. Morbidity. Currently, equations are available in prior art to calculate risk of intraventricular hemorrhaae (a dangerous complication of prematurity associated with brain damage), and neonatal chronic lung disease (an expensive, disabling consequence of extreme prematurity). c. Lengthy of Stav. Illness seven'tv has a powerful influence on length of stav for newborns. The present invention can assist to predict the length of stay and of populations of newborns which is extremelv helpful in projecting workload and occupancv and comparing efficiency of clinical practice styles.

generationusingsimplecharts5.Report and renderedpossiblebyis the svstem of the present invention, to compare current and past performance severityprediction.regardingillness 5. Benchmarking. L : sln2 the reference data gathered by our three research institutions, available in the database and new data which can be generated by the present invention, it is possible to compare performance of any given NICU with all others and with \MCLs of similar type of patient characteristics. This benclunarkina is extremely valable to clinicians, administrators and insurers.

Specificzpplications of SWAP-IN In addition to the"system"implementation described above. SNAP-II can be used independently in research, quality improvement, financial projections and medico- legal risk management.

1. Research. The variety of research applications of SNAP-II has been wide and innovative (see Richardson et al."Neonatal Illness Severity Score: can they predict mortality and morbidity ?" Clinics in Perinatology, 1998; 25 : 591-611).

Applicants are interested in making SLNA-P-II available to legitimate researchers to sustain those innovative applications.

2. Quality Improvement. SNAP-II can serve as an important marker for quality improvement activities. A pattern of unexpectedly ill admissions should prompt review of obstetric practices and pre-admission stabilization. A pattern of death or morbidity in patients with limited risk should prompt review of care practices and clinicians. A pattern of admission of low risk patients should prompt review of staffing and admissions policies. Many of the review processes are mandate by the Joint Commission for the accreditation of hospitals and other organizations and boy mans state regulations. riskmanagement.Severalofthequalityimprovement3.Medico-legal screens noted aboie can also be sued bv ris1 management to select cases for review to identifv and reduce risks of litigation.

54. Financial projections. Current case mix adjusters are crude and retrospective. SN'AP-II offers the possibilitv of an objective, prospective, quantitative measure of expected costs and length of stay. This can be used by hospital systems for budgetina and staffing projections, and by insurers for gauging financial exposure and for gauging financial exposure and for negotiating better reimbursement contracts.

5. Not intended for use in the ethical decision making. Applicants make no claims for the use of SSAP-II or any associated scores in estimating individual mortality risks for the purposes of withdrawing life support. All of applicants' publications have emphasized that there is insufficient certainty in estimates on any individual patient and that such decisions must be made based on clinical jugement and not based on scores.

SCORE FOR PHYSIOLOGY-II(SNAP-IITM)ACUTE Definitions Of Physiologic Variables LOWEST MEAN BLOOD PRESSURE: Lowest mean BP dring the first 12 hours of admission, as recorde in the nursing flow sheet. If only systolic and diastolic are recorded. assume mean BP = diastolic T 1/3 (systolic-diastolic).

LOWEST TEWPERaTURE Lowest body temperature (axillary or rectal but NOT skin probe). This is usually recorde in °F. If recorde in'C, must convert to °F in order to score.

LOWEST pH: Lowest pH dring the first 12 hours of admission. This may be obtained by ABG.

CBG or VBG, and need not be related to the @LOW, @AWP or @HIO blood axasses listed below.

LINKED RESPIRATORY VARIABLES: Thé goal is to identify the three worst arterial blood axasses. These are then used to compute the pAO2/FiO2 ratio assessing oxvgenation status. These three blood tasses are labeled CALOW, CAAWP and @HIO and are recorde as RAW DATA.

@. LOW: Low Blood cas is that with ABSOLUT LOWEST pAO2 during the TIME PERIOD. CBG, VBG, TcPO2 are not acceptable alternatives. Record the following corresponding to the @LOW blood cyans: HighestFiO2expressedaspercent(21%-100%).#FiO2@LOW: <BR> pAO2inmmHg.#pAO2@LOW: @AWP: Lowest pAO2 documente at AWP BLOOD GAS. AWP Blood aas is that with the HIGHEST MEAN AIRWAY PRESSURE. If there are several ABGs at that hiIean Airway pressure, select the one with the worst pAO2 unless that is already recorde as the pAO2CLOW. In that case, select the next lowest. If only one blood gas was obtained, skip this step and proceed to computations. CBG, VBG, TcPO2 are not acceptable alternatives. Record each of the following corresponding to the @AWP blood gas: HighestFiO2expressedaspercent(21%-100%).#FiO2-@AWP: Should correspond to the highest MEAN AIRWAY PRESSURE not already recorde in SLOW. pAO2inmmHg.@pAO2@LOW: @HIO: Lowest pAO2 documente at HIO BLOOD GAS. HIO Blood cas is that with the HIGHEST FiO21. If there are several ABGs at that highest FiO2, select the one with the worst pAO2 unless that is already recorde as the @ LOW or the @AWP pAO2. In that case, select the next lowest. If only one blood eas was obtained, skip this step and proceed to computations. CBG. VBG. TYPO= are not acceptable alternatives. Record each of the following corresponding to the CA, HIO blood gas: HighestFiO2expressedaspercent(21%-100%).#FiO2-@HIO: Should correspond to the hiahest FiO not already recorded in (â LOW or@AWP.

pAO2 tHIO: pAO2 in = Ho.

RATIO:LOWESTpAO2/FiO2 The lowest pAO,/FiO-, ratio is computed using pAO2 in torr, and Fiv2 as percent inspiredoxygen=2)).Usethepairsfrom@LOW,@AWPor(e.g.:80torr/40 % @HIO, and select whichever is lowest.

URINE OUTPUT (CC/KG/HR) : Total Cubic centimeters of urine output dring the first 12 hours of admission divided by BIRTH WEIGHT IN KG, and then divided by 12 HOURS. If notes indicate that some output was lost/unmeasured, then score as 0.

SEIZURE Multiple seizures, confirme or high degree of suspicion. A single seizure or suspecte seizure does not qualify.

ALTERNATIVE PARAMETER MEASUREMENT: Blood pressure, pH, temperature and oxygenation ratio as well as Apgar ratio may be measured, tracked, and transmitted in any alternative conventional manner, either in an analog or digital fashion, as feasible in the DsICU setting. Structured details of such monitors, transducers, analog/digital converters, and measuring systems are not critical to the present invention and are known to those skilled in the art.

Physical conditions such as presence of seizures and urine measurement may be done either automatically without human intervention, or manually, as conditions permit. In any event, measured values of these parameters/physical conditions also can be subjected to the step of optimizing, to choose a desirable type of measurement. In the particular embodiment described herein aboie, the presence of multiple seizures is a potentially significant predictor.

SNAP-II score computation may be done in real time as data items are entered in the database. Data entre for a 21ven patient may be at one or multiple sittings, such that a current SNAP score is computed at each sitting.

Likewise. the present invention may generate real time model predictions based on computed SNAP score (as often as the score is computed). The present invention would thus be linked to reporting systems or other output means for presenting model predictions in real time to physicians, utilization review, hospital administrators, and the like. To that end, output may initially be in the form of prompts for decision (logic) making accordinQ to industry, local or health care insurance standards, or a combination thereof A recommended data entry format for severity of illness evaluation by usine té present invention is shown in the chart below.

SNAP-IIT"^ Severitv of Illness Evaluation Instruction: 1. The most anormal values of each physiologic parameter should not be noted on this sheet.

2. The scoring period from SNAP in the first 12 hours of each patient's admission. The time of the first vital siens will be used as the time of admission. n manuel for details of definitions. Abstractors must adhere strictlv to these definitions.

TABLE 4 I HiEhest I Lowest I SCORE _ ! Mean Blood Pressure i y 3'" :.' I Temperature <. >. dU 1 Serum pH (art/cap) I-11-- 1 1 Arterial Blood pO2/FiO2SCOREpO2 mmHGratioGasespercent (select worst) SCORE Multiple Seizures Present OutputTotalUrine in first Cccc/kg/hrhours SNAP-1111 Score j Birth Weiaht Gestational Age SGA (<')"percentile) Yes No Apaar <7 at 5 minutes Yes Isio SNAPPE-IITM Score

Equivalents While this invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and the scope of the invention as defined by the appende claims. Those skilled in the art will recognize or be able to ascertain usina, no more than routine experimentation, many equivalents to the specific embodiments of the invention described specifically herein. For instance, even though the monitoring of the absence/presence of seizures, and the measurement of urine output are shown in the exemplary embodiment as manual inputs, it is conceivable that urine output measurements and monitoring for seizures can be automate without any human intervention. The automate inputs from the urine output measurements and seizure monitoring can be directly fed into the user interface and/or the unit 105 comprising the sofrivare proram and computer in the schematic Figure 1. Also. instead of the preselected physical conditions of blood pressure, pH. temperature, oxygenation ratio, urine output, and presenceiabsence of multiple seizures, other physical conditions mav be chosen for monitoring, if such choice facilitates a specific study or comparison. Likewise, in lieu of the added measurements on parameters of birth weight, smallness for gestational age, and low Apgar score, other parameters can be chosen to arrive at a specific aspect of illness severity measurement or mortality rate. All such choices of mesurable physical conditions, and different possible methods of measurement including automated. manual, and hvbrid measurements are within the purview of this invention. Such equivalents are intended to be encompassed in the scope of the appende claims.