Release Notes: Acknowledgement Version: 2013 Initial Date: 11/12/2012 ACKNOWLEDGEMENT

Transcription

1 Release Notes: Acknowledgement ACKNOWLEDGEMENT The ORYX Risk Adjustment Guide defines the methods The Joint Commission expects individual performance measurement systems to implement to ensure that their core and non-core measure data are accurate, complete, and timely. Example Acknowledgement: The ORYX Risk Adjustment Guide [Version xx, Month, Year] is the work of The Joint Commission. The manual is periodically updated by The Joint Commission. Users of the ORYX Risk Adjustment Guide must update their software and associated documentation whenever The Joint Commission releases a new version of this manual. ORYX Risk Adjustment Guide

2 Release Notes: Table of Contents Acknowledgement Table of Contents Section 1 Section 2 Section 3 Using the ORYX Risk Adjustment Guide Risk Adjustment Introduction Risk Model Application Steps Examples of Risk Adjustment Data Element Calculations Risk Model Information File Layout Risk Model Information File Example Risk Model Flowchart Example Release Notes Risk Factor Definitions Introduction File Layout of Risk Factor ICD-9 Specifications File Definitions of Risk Factors -- Health Care-Associated Bloodstream Infections in Newborns (PC-04) Release Notes Aggregate-Level Risk Adjustment through Direct Standardization Introduction Application Steps for Aggregate-Level Risk Adjustment using Direct Standardization Stability of the National Maternal Distribution Weight at First Delivery Examples of Calculating Risk-Adjusted Rates -- Perinatal Care Caesarian Section (PC-02) Release Notes i ORYX Risk Adjustment Guide

3 Release Notes: Using the Manual Using the ORYX Risk Adjustment Guide Risk adjustment is a statistical process used to identify and adjust for variation in patient outcomes that stem from differences in patient characteristics (or risk factors) across health care organizations. Depending on the presence of risk factors at the time of health care encounters, patients may experience different outcomes regardless of the quality of care provided by the health care organization. Comparing patient outcomes across organizations without appropriate risk adjustment can be misleading. By adjusting for the risks associated with outcomes of interest, risk adjustment facilitates a more fair and accurate inter-organizational comparison. This manual is intended to provide the technical and statistical requirements that performance measurement systems must implement if they utilizing a national hospital quality measure set that contains measures that are risk-adjusted by the Joint Commission. The following is a brief overview of the information contained within each section of the manual. Section 1 Risk Adjustment This section provides the information related to applying the risk models. This type of risk adjustment is performed on patient-level data and multiple risk factors may be adjusted for. In addition, examples of risk adjustment calculations are included. Section 2 Risk Factor Definitions This section provides the risk factors that can be used in the risk models. An accompanying Excel spreadsheet contains the list of specific diagnosis and procedure codes that are used for each risk factor defined by codes. Section 3 Aggregate-Level Risk Adjustment through Direct Standardization This section provides the information related to risk adjusting an organization s outcome rate based on a characteristic, usually age, of the national population. This type of risk adjustment is performed on aggregate data and typically adjusts for only one risk factor. In addition, examples of risk adjustment calculations are included. An accompanying Excel spreadsheet contains the Direct Standardized information needed to perform the risk adjustment. ORYX Risk Adjustment Guide -i-

4 Risk Adjustment Release Notes: Risk Adjustment Introduction Risk adjustment is a statistical process used to identify and adjust for variation in patient outcomes that stem from differences in patient characteristics (or risk factors) across health care organizations. Depending on the presence of risk factors at the time of health care encounters, patients may experience different outcomes regardless of the quality of care provided by the health care organization. Comparing patient outcomes across organizations without appropriate risk adjustment can be misleading. By adjusting for the risks associated with outcomes of interest, risk adjustment facilitates a more fair and accurate inter-organizational comparison. Below is a general outline of the risk adjustment process being applied to selected national hospital quality measures: ORYX vendors must prepare data for risk adjustment by first identifying the measure population for which risk adjustment is to be used. Relevant binary risk factors must be identified for the measure and recoded to indicate the presence or absence of the risk factor. However, for continuous risk factors, the actual value of the risk factor is used with some modifications. If required risk factors such as sex, age, or admission source are blank or missing, the measurement system must provide an estimated value for each of these missing risk factors. The measurement system must then apply the Risk Models included in the Risk Model Information File, provided by the Joint Commission, to their data to calculate predicted values for each episode of care in the measure population (i.e., numerators and denominators) for risk adjusted measures. These are the predicted values transmitted to The Joint Commission as part of the hospital clinical data in XML format for each of the participating health care organizations. The system must then calculate aggregate each hospital s risk adjustment data elements for the measure (e.g., Risk-Adjusted Rate) and transmit them to the Joint Commission in the aggregate HCO-level data file (X12) for each of the participating health care organizations. Refer to the Aggregate Technical Implementation Guide for further details. The risk model information files will be provided to ORYX vendors that have successfully embedded a national hospital quality measure set that requires risk adjustment for one or more measures. The Joint Commission reserves the right to update or modify risk models as necessary to provide the best and most current models available. It is strongly recommended that ORYX vendors develop a flexible process that allows them to efficiently accommodate modifications to existing risk factor coefficients as well as the addition and/or deletion of risk factors. Modifications may also include changes to risk factor definitions. The only national hospital quality measure currently requiring patient-level risk adjustment is the Health Care-Associated Bloodstream Infections in Newborns (PC-04) outcome measure in the perinatal care measure set. ORYX Risk Adjustment Guide 1-1

5 Release Notes: Risk Adjustment The following section describes in more detail the risk adjustment process that the ORYX vendors must implement. For illustration purposes, a generic measure ABC with generic risk factors Risk1 and Risk2 are referenced throughout this section. There is one risk-adjusted measure that is adjusted at the health care organization aggregated level, Cesarean Section (PC-02). Since the method of calculating the risk-adjusted rate is not the same as the other measures in this chapter, the methodology for the PC-02 measure is explained in the Aggregate-Level Risk Adjustment through Direct Standardization section of this manual. Risk Model Application Steps ORYX vendors must complete the following steps in order to properly risk adjust and submit riskadjusted data to the Joint Commission: Step 1 -- Identify the measure population through Measure Category Assignments. Risk adjusted rate-based measure: Identify the numerator (Measure Category Assignment = E) and the denominator (Measure Category Assignment = D) cases using the information provided in the Measure Information Form (MIF). Risk adjusted continuous variable measure: Identify the number of cases in the measure population (Measure Category Assignment = D). At this time, there are no risk adjusted continuous outcome measures in any of the national hospital quality measure sets. Note: Do not calculate a Predicted Value for a case if it is rejected by front-end edits or is rejected because one or more measures in the measure set evaluates to a Measure Category Assignment = X. Step 2 -- Create risk factors for the measure. Using the Risk Model Information File provided by the Joint Commission, identify all applicable EOC record data elements and the associated risk factor values for each of the EOC records identified in step 1. Risk factors include patient demographic and/or clinical factors, which can influence outcomes of care. Some examples of risk factors include age, sex, and comorbidities such as diabetes or a history of hypertension. As an example, Figure 1 lists the data elements required for risk adjustment of generic measure ABC. Using the data for measure ABC, the performance measurement system must identify the risk factors at the EOC record-level, and create data subsets for each participating hospital. Note: Do not round the individual risk factors when calculating the Predicted Value. Rounding does not occur until the after the Predicted Value is calculated. ORYX Risk Adjustment Guide 1-2

6 Release Notes: Risk Adjustment Figure 1: Data Elements Required for Risk Adjustment Measure Measure ABC ORYX Risk Adjustment Guide 1-3 Risk Adjustment Data Elements -- Age (Calculated from Admission Date and Birthdate) -- ICD-9-CM Principal Diagnosis Code -- ICD-9-CM Other Diagnosis Codes (0 to 17 occurrences) -- Sex A comprehensive list of the risk factors and corresponding definitions is provided in the Risk Factor Definition section of this manual. In addition, an Excel spreadsheet that contains the specific diagnosis and procedure codes for each risk factor will be provided to ORYX vendors when the risk model information file is distributed. Overtime, new patient risk factors may be identified. Both the Risk Factor Definition section of this manual and the Excel spreadsheet will be updated when this occurs. Risk factors can be classified into two types: binary (dichotomous) or continuous. Binary risk factors are assigned a numerical value of one (1) to depict the occurrence or presence of a specific risk factor, or a zero (0) to depict the non-occurrence or absence of the risk factor in the EOC record. For example, binary risk factors can be constructed from the national hospital quality data elements Sex or ICD-9-CM Other Diagnosis Codes. The logic for creating a binary risk factor is illustrated in the following example. Example: Risk1 is the variable name assigned to generic risk factor 1. Perform the following logic for each EOC record: If one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Codes) are one of the identified codes, then Risk1 = 1 (occurrence); else Risk1 = 0 (non-occurrence). Note: Use only the principal and 0 to 17 other diagnosis codes that are sent to The Joint Commission to evaluate the measure. Do not include additional diagnosis codes that may be available in the medical record but are not transmitted in the XML file to The Joint Commission. Continuous risk factors can take on an infinite number ( ) of values, from - to +. They can be entered directly or created from other data elements. For example, the risk factor age is derived from national hospital quality measure data elements Admission Date and Birthdate. Example: AGEINT is the variable name assigned to Integer Age. Perform the following logic for each EOC record: Calculate the integer value of age. Use the month and day portion of Admission Date and Birthdate to yield the most accurate age. Once calculated, use the age in years, expressed as an integer. Do not round up. Risk factor interactions may be used to explain the joint effect of two or more independent variables (risk factors) on a dependent variable (outcome of interest). Interactions occur when two or more risk factors not only have separate effects but also have combined effects on the outcome

7 Release Notes: Risk Adjustment of interest. In other words when an interaction is significant, it is implied that the effect on the outcome interest of one factor depends on the level of another factor. For example, the rate of AMI mortalities increases with age, however this rate is increased more if the patient is male. This implies a two-way interaction of age and sex. In the measure risk adjustment process, the interaction will be depicted by two or more risk factors connected by an underscore as in AGEINT_SEXR, indicates an interaction of integer value of age and sex. For calculation purposes, the risk factor AGEINT is multiplied by the variable SEXR. Example: If an interaction between age and sex has a coefficient of , the age is given as 55, and the sex is female, then the calculation for this interaction would be: AGEINT = 55 and SEXR = 1 (female) AGEINT_SEXR = (coefficient * AGEINT * SEXR) = ( * 55 * 1) = Step 3 -- Managing missing risk factor data. Risk adjusting national hospital quality measures without assessing the extent of missing data for risk factors may result in the transmission of biased data. In particular, if missing data occur in a nonrandom fashion, the data analysis can be misleading. Several statistical methods exist that can be used to handle missing risk factor data. ORYX vendors are required to replace any missing risk factor data with estimated (or imputed) data using the procedures described below, regardless of the missing risk factor percentages. If an EOC is a member of multiple measures, the missing risk factor data may be assigned a different estimated value for each measure. Replacement of missing risk factors occurs after each EOC s Measure Category Assignment has been determined. 1. Missing data percentage: Calculate the number of EOC records per measure per hospital per month, that include one or more missing risk factors (i.e., contained in the Risk Model Information File) for the measure. ORYX vendors need to refer to the ORYX Data Quality Manual and Aggregate Technical Implementation Guide for additional information. 2. Continuous risk factor: Calculate an estimated value using the arithmetic average (rounded to six decimal places) of all organizations denominator data (i.e. category Ds and Es) available for the measure. Age: calculate and retain only the integer portion. This estimated value should be inserted for all missing data occurrences of that risk factor for all hospitals. Birth Weight: If birth weight is missing, then the ICD-9 codes should be used to code the Birth Weight risk factors. If a valid birth weight is present and it disagrees with the ICD-9 code for the birth weight risk factor, then the birth weight should be used to code the risk factor. 3. Binary risk factor: The estimated value used is the most frequently occurring response (or mode) for all organizations data available for the measure. This estimated value ORYX Risk Adjustment Guide 1-4

8 Release Notes: Risk Adjustment should be inserted for all missing data occurrences of the risk factor in question for all hospitals. When a tie occurs in the frequencies, use the baseline of 0 as the estimated value. It is very important to note that missing value replacements do not apply to clinical risk factors because these are derived from ICD-9-CM Codes. If an ICD-9-CM Other Diagnosis Code field, used to identify risk factor is not populated, it is assumed that no risk factor was associated with the patient in that field. It is therefore not likely at all that every ICD-9-CM Other Diagnosis Code data element field will be populated. Step 4 -- Calculate the EOC predicted values using risk model information file Calculate the Predicted Values for each EOC record using the risk model information provided by the Joint Commission. The Joint Commission will specify the risk model equation type in the Risk Model Information File. ORYX vendors must use the risk model equation to calculate predicted values. For example, Measure ABC uses logistic regression as the risk model equation type, since it is a rate-based proportion measure. Continuous variable measures (to be considered in the future) will utilize linear regression methods. Additionally, the Risk Model Information File will contain the measure s risk factor variables and related regression coefficients. See Figure 2 for information concerning the layout of the Risk Model Information File. See Figure 3 for an example of a Risk Model Information File later in this section. Regression coefficients are a common component of risk models and are defined as a weight or number that indicates the associated relationship between a dependent variable (outcome of care) and an independent variable (risk factor). All calculations of Predicted Values are computed at the EOC record-level using measure population data. The Predicted Values are calculated by first multiplying regression coefficients by an appropriate risk factor value. If the risk factor is binary, then the value indicates the presence or absence of a factor. If the risk factor is continuous, then the risk factor is an appropriate value specifying a level of the variable. A more detailed explanation of this calculation process is given below. For each risk-adjusted measure the Predicted Value is transmitted along with the EOC record to The Joint Commission in the Hospital Clinical Data (HCD) XML file. The following equation calculates the Predicted Value (y) [i.e., probability], rounded to 8 decimal positions, for rate-based measures: 1 Predicted Value(y) = -( B0+B 1* X Bn* X n ) 1+e Where: o e V is the exponential function [also may be written as exp(v)] that is the value of the intercept parameter generated by each model; o B0...Bn are the Joint Commission provided regression coefficients; and ORYX Risk Adjustment Guide 1-5

9 Release Notes: Risk Adjustment o X1...Xn are the calculated risk factor values. The exponential value (e) is truncated to 8 decimal points for calculation purposes. The exponential value (e) is an irrational number that is a universal constant and is the base of natural logarithms. The exponential value should be used in the calculations rather than the exponential function. e 1 =exp(1)= Step 5 -- Calculate the HCO-level aggregate risk adjustment data elements. By aggregating the Predicted Value for each EOC, calculate each hospital s risk adjusted data elements (e.g., Risk-Adjusted Rate) that are transmitted to The Joint Commission in the HCO-level data file (X12). ORYX vendors need to refer to the ORYX Data Quality Manual and Aggregate Technical Implementation Guide for additional details, formulas, and transmission specifications. Note that once the X12 file is retired that The Joint Commission will make this calculation and this will no longer be submitted ORYX Risk Adjustment Guide 1-6

10 Release Notes: Risk Adjustment Examples of Calculating Predicted Values The following examples demonstrate steps used to calculate the Predicted Value (y) for an EOC record in a rate-based measure population. Example 1: Measure ABC Suppose a patient has the following risk factors. This patient s predicted mortality probability can be calculated using a hypothetical risk model for a generic measure. EOC Record Data Element Value Risk Factor Name (Risk factors included in the model) Risk Factor Value Sex = female SEXR (male = 0; female = 1) X 1 =1 Admission Date Birthdate = 75 AGEINT = 75 X 2 =75 (age) ICD-9-CM Principal Diagnosis Risk1 (Yes=1, No=0) X 4 =1 Code = ICD-9-CM Other Diagnosis Code Risk2(Yes=1, No=0) X 5 =1 = V45.81 ICD-9-CM Other Diagnosis Code Risk3 (Yes=1, No=0) X 9 =0 = ICD-9-CM Other Diagnosis Code = V45.82 Risk4 (Yes=1, No=0) X 10 =1 Example Regression Coefficients Provided by the Joint Commission for measure ABC : B 0 = Intercept Constant B 1 = Coefficient for SEXR (X 1 ) B 2 = Coefficient for AGEINT (X 2 ) B 3 = Coefficient for Risk1 (X 3 ) B 4 = Coefficient for Risk2 (X 4 ) B 5 = Coefficient for Risk3 (X 5 ) B 6 = Coefficient for Risk4 (X 6 ) ORYX Risk Adjustment Guide 1-7

11 Release Notes: Risk Adjustment Using the risk factor values X 1, X 2, X 3... X 10 in the example EOC record, the Predicted Value (y) and the regression coefficients from the example above can be calculated as follows: 1 Predicted Value (y) = 1 e V, Where V= B 0 + B 1 X B n X n Therefore, V in this example is calculated to be the sum of the following operations: V = *(1) *(75) *(1) *(1) *(0) *(1) = The Predicted Value (y) for the EOC in this example calculates to: 1 (y) = (0.3321) 1 e 1 1 ( ) = = , which rounded to 8 digits is After the predicted value is calculated for every EOC record in the measure population, the HCOlevel risk-adjusted data elements (e.g., Risk-Adjusted Rate) can be calculated by aggregating EOC-level predicted values for each hospital. ORYX Risk Adjustment Guide 1-8

12 Release Notes: Risk Adjustment Figure 2: Risk Model Information File Layout Field Format Content Name Quarter N6 Specifies a combination of the Year and Quarter to which the risk factors and model should be applied (e.g., references 3 rd quarter 2012) Measure_ID N10 Contains the unique Performance Measure Identifier (used for ORYX data transmission) for the measure to which the risk factor applies Eq_Type N1 Identifies which risk model equation is to be used for the Predicted Value calculation 1 = If logistic regression model, i.e., Predicted Value ( y) 1 e 1 ( B0 B1 X1 B2 X 2... B n X n ) 2 = If multiple regression model with transformation of dependant variable, i.e., Predicted Value (y) =e B0+B1*X1+B2*X2+...+Bn*Xn 3 = If multiple regression model without transformation, i.e., Predicted Value (y) =B 0 +B 1 *X 1 +B 2 *X B n *X n Where B 0...B n are the Joint Commission provided regression coefficients, X 1...X n are the created risk factor values, and e is the exponential function with e= Factor_ID AN12 Identifies the risk factor by a unique name N = Not applicable (to be used for constant term/intercept B 0 ) Factor_Status N1 Identifies whether the risk factor is existing or new 1 = existing (i.e., risk factor as defined in the current listing) 2 = new (i.e., changed risk factor as defined in the new listing, or a new risk factor) 3 = not applicable (to be used for constant term/intercept B 0 ) Factor_Type AN1 Identifies whether the risk factor is continuous or binary C = Risk factor is continuous B = Risk factor is binary N = Not applicable (to be used for constant term/intercept B 0 ) Short Name AN40 Briefly describes the risk factor. Refer to the Risk Factor Definition section in this manual for a full definition of each risk factor. Coefficients N3.6 Contains regression coefficients, accurate to 4 decimal places Must support a leading negative sign for negative numbers, e.g., Positive = and Negative = ORYX Risk Adjustment Guide 1-9

13 Release Notes: Risk Adjustment Figure 3: Risk Model Information File Example This sample is for a generic risk model information file and does not reflect any current risk model. The current version of the complete risk model information file layout is provided to ORYX vendors on the Performance Measurement System Extranet Track (PET). Quarter Measure ID Equation Type Factor ID Factor Status Factor Type Short Name Coefficient yyyyqq N 3 N Constant term yyyyqq AGET C Age truncated yyyyqq SEXR 1 B Sex =Male or Female yyyyqq Risk1 1 B Risk Factor yyyyqq Risk2 1 B Risk Factor yyyyqq Risk3 1 B Risk Factor yyyyqq Risk4 1 B Risk Factor ORYX Risk Adjustment Guide 1-10

14 Release Notes: Risk Adjustment Risk Model Example ABC: Inpatient Mortality Numerator: Inpatient mortality of AMI patients. Denominator: AMI patients. ORYX Risk Adjustment Guide 1-11

15 ORYX Risk Adjustment Guide 1-12 Release Notes: Risk Adjustment

16 Release Notes: Risk Adjustment 2013 Release Notes Items highlighted in yellow have been modified or added from the previous version of the manual. These release notes detail the changes that have been made to this document since its initial version was provided to ORYX vendors on 11/12/2012. Items that have been modified or added are highlighted in blue. Date of Original Information Modified Information Reason Modification 11/12/2012 Risk adjustment examples have become generic and not examples of specific measures. To eliminate the need to modify this document whenever a risk-adjusted measure has been added, retired, or changed. 11/12/2012 Removed information concerning AMI-9. AMI-9 is retired. ORYX Risk Adjustment Guide 1-13

17 Risk Factor Definitions Release Notes: Risk Factor Definitions Introduction This section provides the comprehensive list of risk factors that may be used in the risk model for each outcome national quality measure. The information includes the risk factor variable name used in the risk model, the risk factor s short name, the type of risk factor (binary or continuous), and the definition of the risk factor. Risk factors that are defined by ICD-9 codes are denoted as such, but the actually ICD-9 codes are not contained in this document. Refer to the Excel spreadsheet that accompanies this document for the specific diagnosis and procedure codes that are appropriate for each risk factor. File Layout of Risk Factor ICD-9 Specifications File The file format is Excel (*.xls). ICD-9 Specifications File Layout: Data Element Name Definition 1. Measure Short Name (Set Measure The name and ID of the measure that the risk factor is associated to. ID#) 2. Risk Factor Variable Name The risk factor variable name in the Excel file matches the variable name provided below in the definition of the risk factors. 3. Risk Factor Short Name The risk factor short name in the Excel file matches the short name provided below in the definition of the risk factors. 4. ICD-9-CM Code -- This column contains both diagnosis and procedure codes. Use column 6 (Type) in conjunction with this data to determine the type of code. -- This column is blank if the associated risk factor does not use ICD- 9 codes to define it. Refer to this section of the manual to determine how to define these risk factors. -- Several risk factors use ICD-9 codes and other data elements to define them. This section of the manual and the ICD-9 Specifications File must be used in unison to ensure the risk factors are properly coded. -- The ICD-9 codes do not contain decimal points. The decimal points are implied. 5. Description This is the description of the ICD-9 code contained in the ICD-9-CM Code column. 6. Type: D = Diagnosis, P = Procedure This column further defines the data contained in the ICD-9-CM Code column (column 4). -- D = Diagnosis Code -- P = Procedure Code -- Blank means the ICD-9-CM Code column is blank Note: In previous versions, this was column 11. The change is occurring with 3Q12 data being transmitted by 1/31/2013. ORYX Risk Adjustment Guide 2-1

18 Release Notes: Risk Factor Definitions Definitions of Risk Factors Perinatal Care Measure Set Health Care-Associated Bloodstream Infections in Newborns (PC-04) Risk Factor Risk Factor Type of Definition Variable Name Short Name Risk Factor SEXR Sex = M or F Binary If Sex = F, then SEXR = 1; If Sex = M, then SEXR = 0; If Sex = U, stop evaluating the case. No predicted value will be assigned for this case and it will not be included in the evaluation to calculate the Risk-Adjusted Rate. RF369 New born with birth weight between 1000 and 1249 grams Binary If Birth Weight is >= 1000 grams and Birth Weight is <= 1249 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF369 = 1 ; else RF369 = 0 RF371 RF372 RF373 Congenital Gastrointestinal Anomaly Congenital Cardiovascular Anomaly Congenital Other Anomaly Binary Binary Binary Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. If one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF371 = 1 ; else RF371 = 0 If one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF372 = 1 ; else RF372 = 0 If one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF373 = 1 ; else RF373 = 0 RF374 Newborn transfers in Binary If ICD-9-CM Principal Diagnosis Code is not on Table in Appendix A of the Specifications Manual, then RF374 = 1; else RF374 = 0 RF375 RF376 Newborn transfers out or died New Born with Birth Weight Between 500 and 749 grams Binary Binary If Discharge Disposition= 6 (Expired) or Discharge Disposition = 4 (Discharged/transferred to a Acute Care Facility), then RF375 = 1; else RF375 = 0 If Birth Weight is >= 500 grams and Birth Weight is <= 749 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF376 = 1 ; else RF376 = 0 Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. ORYX Risk Adjustment Guide 2-2

19 Release Notes: Risk Factor Definitions Risk Factor Variable Name RF377 Risk Factor Short Name New Born with Birth Weight Between 750 and 999 grams Type of Risk Factor Binary Definition If Birth Weight is >= 750 grams and Birth Weight is <= 999 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF377 = 1 ; else RF377 = 0 RF378 New Born with Birth Weight Between 1250 and 1499 grams Binary Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. If Birth Weight is >= 1250 grams and Birth Weight is <= 1499 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF378 = 1 ; else RF378 = 0 RF379 New Born with Birth Weight Between 1500 and 1749 grams Binary Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. If Birth Weight is >= 1500 grams and Birth Weight is <= 1749 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF379 = 1 ; else RF379 = 0 RF380 New Born with Birth Weight Between 1750 and 1999 grams Binary Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. If Birth Weight is >= 1750 grams and Birth Weight is <= 1999 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF380 = 1 ; else RF380 = 0 RF381 New Born with Birth Weight Between 2000 and 2499 grams Binary Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. If Birth Weight is >= 2000 grams and Birth Weight is <= 2499 grams OR one or more diagnosis codes (ICD-9-CM Principal Diagnosis Code or ICD-9-CM Other Diagnosis Code) are one of the identified codes, then RF381 = 1 ; else RF381 = 0 Note: If a valid Birth Weight is present and it disagrees with the ICD-9 code for the birth weight, then use the Birth Weight to code the risk factor. ORYX Risk Adjustment Guide 2-3

20 Release Notes: Risk Factor Definitions 2013 Release Notes Items highlighted in yellow have been modified or added from the previous version of the manual. These release notes detail the changes that have been made to this document since its initial version was provided to ORYX vendors on 11/12/2012. Items that have been modified or added are highlighted in blue. Date of Modification Original Information Modified Information Reason 11/12/2012 Removed columns 6 through 10 from the file layout for the Risk Factor ICD-9 Specifications File. This moved the Type (of data) column up from #11 to #6. 11/12/2012 Removed information concerning AMI-9 from the document. 11/12/2012 Added SEXR as a risk factor to PC /12/2012 Added new risk factors to PC-04 RF376, RF377, RF378, RF379, RF380, and RF /12/2012 Modified existing risk factors for PC-04 RF374 and RF /12/2012 Removed risk factors no longer being used from PC-04 - RF368 and RF370. Columns 6 through 10 are not used when calculating the risk model. AMI-9 is retired. Modification occurred to mirror data element changes in the Specifications Manual. ORYX Risk Adjustment Guide 2-4

21 Risk Factor ICD-9 Code Specifications Effective starting with 7/1/2012 Discharges type: D=Diag Measure Risk Factor Risk Factor Short Name ICD-9- CM Code Description, P=Proc HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 4240 Mitral valve disorder D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 4241 Aortic valve disorder D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 4251 Hypertr obstr cardiomyop D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 4253 Endocard fibroelastosis D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7450 Common truncus D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Compl transpos great ves D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Double outlet rt ventric D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Correct transpos grt ves D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Transpos great vess NEC D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7452 Tetralogy of fallot D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7453 Common ventricle D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Endocard cushion def NOS D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Ostium primum defect D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Endocard cushion def NEC D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Cong pulmon valv atresia D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Pulmonary valve anom NEC D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7461 Cong tricusp atres/sten D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7462 Ebstein's anomaly D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7463 Cong aorta valv stenosis D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7464 Cong aorta valv insuffic D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7465 Congen mitral stenosis D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7466 Cong mitral insufficienc D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly 7467 Hypoplas left heart synd D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Cong subaortic stenosis D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Cor triatriatum D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Infundib pulmon stenosis D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Obstruct heart anom NEC D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Coronary artery anomaly D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Coarctation of aorta D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Interrupt of aortic arch D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Anomalies of aortic arch D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Aortic atresia/stenosis D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Cong anom of aorta NEC D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Great vein anomaly NOS D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Tot anom pulm ven connec D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Part anom pulm ven conn D HealthCare-Associated Bloodstream Inf RF372 Congenital Cardiovascular Anomaly Great vein anomaly NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 5602 Volvulus of intestine D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7503 Cong esoph fistula/atres D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7504 Esophageal anomaly NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7505 Cong pyloric stenosis D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7507 Gastric anomaly NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7508 Upper GI anomaly NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7509 Upper GI anomaly NOS D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7510 Meckel's diverticulum D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7511 Atresia small intestine D Page 1 of 3

22 Risk Factor ICD-9 Code Specifications Effective starting with 7/1/2012 Discharges type: D=Diag Measure Risk Factor Risk Factor Short Name ICD-9- CM Code Description, P=Proc HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7512 Atresia large intestine D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7513 Hirschsprung's disease D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7514 Intestinal fixation anom D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7515 Intestinal anomaly NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly Biliary & liver anom NOS D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly Biliary atresia D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly Biliary & liver anom NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7517 Pancreas anomalies D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7518 Anom digestive syst NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7519 Anom digestive syst NOS D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly Congn anoml abd wall NOS D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly Congn anoml abd wall NEC D HealthCare-Associated Bloodstream Inf RF371 Congenital Gastrointestinal Anomaly 7711 Cong cytomegalovirus inf D HealthCare-Associated Bloodstream Inf RF373 Congenital Other Anomaly 7596 Hamartoses NEC D HealthCare-Associated Bloodstream Inf RF373 Congenital Other Anomaly 7765 Congenital anemia D HealthCare-Associated Bloodstream Inf RF373 Congenital Other Anomaly 7780 Hydrops fetalis no isoim D HealthCare-Associated Bloodstream Inf RF376 New Born with Birth Weight Between 500 and 749 gra LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF376 New Born with Birth Weight Between 500 and 749 gra LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF376 New Born with Birth Weight Between 500 and 749 gra FETAL MALNUTR G D HealthCare-Associated Bloodstream Inf RF376 New Born with Birth Weight Between 500 and 749 gra FET GROWTH RET G D HealthCare-Associated Bloodstream Inf RF376 New Born with Birth Weight Between 500 and 749 gra EXTREME IMMATUR GD HealthCare-Associated Bloodstream Inf RF376 New Born with Birth Weight Between 500 and 749 gra PRETERM NEC G D HealthCare-Associated Bloodstream Inf RF377 New Born with Birth Weight Between 750 and 999 gra LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF377 New Born with Birth Weight Between 750 and 999 gra LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF377 New Born with Birth Weight Between 750 and 999 gra FETAL MAL G D HealthCare-Associated Bloodstream Inf RF377 New Born with Birth Weight Between 750 and 999 gra FET GROWTH RET G D HealthCare-Associated Bloodstream Inf RF377 New Born with Birth Weight Between 750 and 999 gra EXTREME IMMATUR GD HealthCare-Associated Bloodstream Inf RF377 New Born with Birth Weight Between 750 and 999 gra PRETERM NEC G D HealthCare-Associated Bloodstream Inf RF369 New Born with Birth Weight Between 1000 and 1249 g LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF369 New Born with Birth Weight Between 1000 and 1249 g LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF369 New Born with Birth Weight Between 1000 and 1249 g FETAL MAL G D HealthCare-Associated Bloodstream Inf RF369 New Born with Birth Weight Between 1000 and 1249 g FET GRWTH RET GD HealthCare-Associated Bloodstream Inf RF369 New Born with Birth Weight Between 1000 and 1249 g EXTREME IMMAT GD HealthCare-Associated Bloodstream Inf RF369 New Born with Birth Weight Between 1000 and 1249 g PRETERM NEC G D HealthCare-Associated Bloodstream Inf RF378 New Born with Birth Weight Between 1250 and 1499 g LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF378 New Born with Birth Weight Between 1250 and 1499 g LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF378 New Born with Birth Weight Between 1250 and 1499 g FETAL MAL G D HealthCare-Associated Bloodstream Inf RF378 New Born with Birth Weight Between 1250 and 1499 g FET GRWTH RET GD HealthCare-Associated Bloodstream Inf RF378 New Born with Birth Weight Between 1250 and 1499 g EXTREME IMMAT GD HealthCare-Associated Bloodstream Inf RF378 New Born with Birth Weight Between 1250 and 1499 g PRETERM NEC G D HealthCare-Associated Bloodstream Inf RF379 New Born with Birth Weight Between 1500 and 1749 g LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF379 New Born with Birth Weight Between 1500 and 1749 g LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF379 New Born with Birth Weight Between 1500 and 1749 g FETAL MAL G D HealthCare-Associated Bloodstream Inf RF379 New Born with Birth Weight Between 1500 and 1749 g FET GRWTH RET GD HealthCare-Associated Bloodstream Inf RF379 New Born with Birth Weight Between 1500 and 1749 g EXTREME IMMAT GD HealthCare-Associated Bloodstream Inf RF379 New Born with Birth Weight Between 1500 and 1749 g PRETERM NEC G D Page 2 of 3

23 Risk Factor ICD-9 Code Specifications Effective starting with 7/1/2012 Discharges type: D=Diag Measure Risk Factor Risk Factor Short Name ICD-9- CM Code Description, P=Proc HealthCare-Associated Bloodstream Inf RF380 New Born with Birth Weight Between 1750 and 1999 g LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF380 New Born with Birth Weight Between 1750 and 1999 g LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF380 New Born with Birth Weight Between 1750 and 1999 g FETAL MALNUTR G D HealthCare-Associated Bloodstream Inf RF380 New Born with Birth Weight Between 1750 and 1999 g FET GRWTH RET GD HealthCare-Associated Bloodstream Inf RF380 New Born with Birth Weight Between 1750 and 1999 g EXTREME IMMAT GD HealthCare-Associated Bloodstream Inf RF380 New Born with Birth Weight Between 1750 and 1999 g PRETERM NEC G D HealthCare-Associated Bloodstream Inf RF381 New Born with Birth Weight Between 2000 and 2499 g LT-FOR-DATES G D HealthCare-Associated Bloodstream Inf RF381 New Born with Birth Weight Between 2000 and 2499 g LT-DATE W/MAL G D HealthCare-Associated Bloodstream Inf RF381 New Born with Birth Weight Between 2000 and 2499 g FETAL MALNUTR G D HealthCare-Associated Bloodstream Inf RF381 New Born with Birth Weight Between 2000 and 2499 g FET GRWTH RET GD HealthCare-Associated Bloodstream Inf RF381 New Born with Birth Weight Between 2000 and 2499 g EXTREME IMMAT GD HealthCare-Associated Bloodstream Inf RF381 New Born with Birth Weight Between 2000 and 2499 g PRETERM NEC G D HealthCare-Associated Bloodstream Inf RF374 Newborn Transfers In HealthCare-Associated Bloodstream Inf RF375 Newborn Transfers Out or Died Page 3 of 3

24 Aggregate-Level Risk Adjustment through Direct Standardization Release Notes: Aggregate-Level Risk Adjustment through Direct Standardization Introduction The risk adjustment methodology used to date by The Joint Commission has used patient-level data to adjust for variation in patient outcomes that stem from differences in patient characteristics (or risk factors) across health care organizations. The expected risk-adjusted rate for an organization is determined based on the patient characteristics of the organization, a method known as indirect standardization since the national rate is adjusted to the organization. Indirect standardization commonly adjusts an organization s rate based on three or more risk factors present in its patient population. In indirect standardization, one calculates the number of outcomes that would have been expected if the outcome rates from the national population had applied in the organization s population. Another method of standardization, known as direct standardization, adjusts an organization s outcome rate based on a characteristic, usually age, of the national population. Direct standardization is based on aggregate data and typically adjusts for only one risk factor. The cesarean section measure in the Perinatal Care measure set (PC-02) uses the direct standardization method to risk-adjust each organization s observed measure rate. For this measure, one applies the national maternal distribution weight at first delivery to an organization s aggregated measure population, by weighting the observed cesarean section rates for each age group according to their national frequency. They are then summed to give the adjusted rate. The adjusted (or expected) rate is then interpreted as what the cesarean section rate would be expected to be if the organization performed at the national rate for each age group. The procedure given below describes how the direct standardization risk-adjusted rate is calculated for an organization. Application Steps for Aggregate-Level Risk Adjustment using Direct Standardization This process is only applicable to proportion measures. ORYX vendors must complete the following steps in order to properly risk adjust and submit aggregate-level direct standardization riskadjusted data to the Joint Commission in the HCO-level data file. Note that once the X12 file is retired that The Joint Commission will make this calculation and this will no longer be submitted. Step 1 -- Identify the measure population through Measure Category Assignments. Identify the numerator (Measure Category Assignment = E) and the denominator (Measure Category Assignment = D) cases using the information provided in the Measure Information Form (MIF). Step 2 -- Within each age stratum, calculate the count of numerator and denominator cases. Aggregate the number of numerator and denominator cases within each age category. The number of numerator cases is the total number of category E s within an age category and the number of denominator cases is sum of the total number of category D s and E s. ORYX Risk Adjustment Guide 3-1

25 Release Notes: Aggregate-Level Risk Adjustment through Direct Standardization Step 3 -- Within each age stratum, calculate the observed measure rates. Within each age stratum, the observed measure rate is the count of numerator cases divided by the count of denominator cases. Step 4 -- Within each age stratum, multiply the observed measure rate by the corresponding National Maternal Distribution Weight at First Delivery to create the weighted measure rates. The Joint Commission will provide the national age distribution weights in the Direct Standardization Information File, available on PET. Step 5 -- Sum up the weighted measure rates over all the age strata to calculate the HCO-level aggregate risk adjustment data elements. Aggregate the weighted measure rates over age strata to calculate each hospital s risk adjusted data element (Risk-Adjusted Rate) that is transmitted to The Joint Commission in the HCO-level (X12) data file. ORYX vendors need to refer to the ORYX Data Quality Manual and ORYX Technical Information Guide for additional details, formulas, and transmission specifications. Note that once the X12 file is retired that The Joint Commission will make this calculation and this will no longer be submitted. Stability of the National Maternal Distribution Weight at First Delivery As depicted below, the National Maternal Distribution Weight at First Delivery has been relatively stable over the last 4 years, even as the overall birth population has decreased. Proportion of US Births in Each Age Band for First Births Population National Maternal Distribution Weight at First Delivery Population National Maternal Distribution Weight at First Delivery Under 15 6, , , , , , , , , , , , , , , , Total 1,696,989 1,603,832 The data contained in the above table is from Table 2. Births, by age of mother, live-birth order, and race of mother: United States: 2010 Martin JA, Hamilton BE, Ventura SJ, et al. Births: Final data for National vital statistics reports; vol 61 no 1. Hyattsvile, MD: National Center for Health Statistics. August, ORYX Risk Adjustment Guide 3-2