December 15, 2005
The Kids' Inpatient Database (KID) is one of a family of databases and software tools developed as part of the Healthcare Cost and Utilization Project (HCUP), a Federal-State-Industry partnership sponsored by the Agency for Healthcare Research and Quality (AHRQ).
The KID is a unique and powerful nationwide database of hospital inpatient stays for children. It is a sample of pediatric discharges (age 20 or less at admission) from community, non-rehabilitation hospitals from states participating in HCUP. The HCUP KID team designed the database to permit researchers to study a broad range of conditions and procedures related to hospitalizations of children. Researchers and policy makers can use the KID to identify, track, and analyze national trends in hospital utilization, access, charges, quality, and outcomes for children.
This report describes the KID sample and weights, summarizes the contents of the 2003 KID, and discusses data analysis issues. Previous KID releases covered 1997 and 2000. This document highlights cumulative information for all previous years to provide a longitudinal view of the database. We have enhanced the nationwide representation of the sample by incorporating data from additional HCUP State Partners. The 2003 KID includes data from 36 states, nine more than the 2000 KID.
Design Considerations
The overall design objective was to select a sample of pediatric discharges that accurately represents the target universe of U.S. community, non-rehabilitation hospitals. Moreover, this sample was to be geographically dispersed, yet drawn exclusively from hospitals in states that participate in HCUP and agree to contribute to the KID.
It should be possible, for example, to estimate Diagnosis Related Group (DRG)-specific average lengths of stay across all U.S. hospitals using weighted average lengths of stay, based on averages or regression coefficients calculated from the KID. Ideally, relationships among outcomes and their correlates estimated from the KID should accurately represent all U.S. hospitals. However, the 2003 KID includes data from only 36 states. Therefore, it is advisable to verify your estimates against other data sources, if available. For example, the National Hospital Discharge Survey (http://www.cdc.gov/nchs/about/major/hdasd/nhds.htm) can provide benchmarks against which to check your national estimates for hospitalizations with more than 5,000 cases.
The KID Comparison Report assesses the accuracy of KID estimates. The most recent report is available on the KID Documentation CD-ROM and provides a comparison of the 1997 KID with other data sources. The updated report for the current KID is expected to be available on the HCUP User Support Website (http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp) in late January, 2006.
Changes to Sampling and Weighting Strategy Beginning with the 2000 KID
We use the HCUP Nationwide Inpatient Sample (NIS) community hospital universe and strata definitions for the KID. We revised some of the NIS hospital universe and strata definitions for 1998 and subsequent data years, and we used these revised definitions beginning with the 2000 KID. These changes included:
A full description of the evaluation and revision of the NIS sampling strategy for 1998 and subsequent data years can be found in the special report, Changes in NIS Sampling and Weighting Strategy for 1998. This document is available on the 2003 KID Documentation CD-ROM and on the HCUP User Support Website at http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp.
Sampling Frame
The KID sampling frame included all pediatric discharges from community, non-rehabilitation hospitals in the HCUP State Inpatient Databases (SID) that could be matched to the corresponding American Hospital Association (AHA) survey data (subject to state-specific restrictions). Beginning with the 2000 KID, pediatric discharges were defined as having an age at admission of 20 or less. This differs from the 1997 KID, which included discharges with an admission age of 18 or less. Discharges with missing, invalid, or inconsistent ages were excluded.
Sampling Procedure
The KID includes a sample of pediatric discharges from all hospitals in the sampling frame. For the sample, we stratified the pediatric discharges by uncomplicated in-hospital birth, complicated in-hospital birth, and pediatric non-birth. To further ensure an accurate representation of each hospital’s pediatric case-mix, we also sorted the discharges by state, hospital, DRG, and a random number within each DRG. We then used systematic random sampling to select 10 percent of uncomplicated in-hospital births and 80 percent of other pediatric cases from each frame hospital.
Discharge Weights
To obtain national estimates, we developed discharge weights using the AHA universe as the standard. For the weights, we post-stratified hospitals on six characteristics contained in the AHA hospital files. These were the same characteristics used to define the NIS sampling strata, with the addition of an additional stratum for freestanding children’s hospitals. Hospital stratification variables were defined as follows:
We merged any stratum containing fewer than two frame hospitals, 30 uncomplicated births, 30 complicated births, and 30 non-birth pediatric discharges sampled with an "adjacent" stratum containing hospitals with similar characteristics. We created the discharge weights by hospital stratum in proportion to the number of AHA newborns for newborns and in proportion to the total number of non-newborn AHA discharges for non-newborns.
Weight Data Elements
To produce nationwide estimates, use the discharge weights to extrapolate sampled discharges in the Core file to the discharges from all U.S. community, non-rehabilitation hospitals. For the 2003 KID, use DISCWT to calculate nationwide estimates for all analyses. For the 2000 KID, use DISCWT to create nationwide estimates for all analyses except those that involve total charges, and use DISCWTCHARGE to create nationwide estimates of total charges. For the 1997 KID, use DISCWT_U for all analyses.
AHRQ obtained agreements from 36 HCUP State Partners to participate in the 2003 KID. More than 90% of the hospital universe is included in the sampling frame for 27 of these states. Texas, supplied data from only 73% of the state’s hospitals because some Texas hospitals, mostly small rural facilities, are exempt from statutory reporting requirements. Minnesota supplied data from only 88% of the state’s hospitals to HCUP because a few Minnesota hospitals do not participate in the project. There are no apparent significant differences between the characteristics of participating and non-participating Minnesota hospitals. We omitted from the sampling frame 32 Michigan hospitals that did not report total charges, leaving 68% of Michigan hospitals in the frame. Six State Partners - Connecticut, Georgia, Hawaii, South Carolina, South Dakota, and Virginia - imposed sampling restrictions that limited the percentage of state hospitals included in the frame to between 54 and 88 percent. (Restrictions from other states did not have an appreciable effect on the percentage of hospitals in the sampling frame.)
Although pediatric discharges from hospitals in each region are selected for the KID, the comprehensiveness of the sampling frame varies by region. The percentage of hospitals included in the sampling frame is highest in the Midwest (89.0%) and in the West (78.0%), while figures are lower for the Northeast (63.9%) and the South (56.9%).
Because the KID sampling frame has a disproportionate representation of the more populous states and contains hospitals with more annual discharges, its comprehensiveness in terms of discharges is higher. The percentage of the regional population in KID states ranges from 99.0% in the Midwest to 74.9% in the Northeast. Overall, the states in the 2003 KID include an estimated 86.5% of the entire U.S population, up from 76.2% in the 2000 KID.
The final 2003 KID sample included 2,984,129 discharges of children from 3,438 hospitals, drawn from 36 frame states representing each region of the United States. The 2003 KID is larger than the 2000 KID across several dimensions:
Missing Values
Missing data values can compromise the quality of estimates. If the outcome for discharges with missing values is different from the outcome for discharges with valid values, then sample estimates for that outcome will be biased and will not accurately represent the discharge population. Also, when estimating totals for non-negative variables with missing values, sums would tend to be underestimated because the cases with missing values would be omitted from the calculations. Several techniques are available to help overcome this bias. One strategy is to impute acceptable values to replace missing values. Another strategy is to use sample weight adjustments to compensate for missing values. Descriptions of such data preparation and adjustment are outside the scope of this report; however, it is recommended that researchers evaluate and adjust for missing data, if necessary.
Variance Calculations
It may be important for researchers to calculate a measure of precision for some estimates based on the KID sample data. Variance estimates must take into account both the sampling design and the form of the statistic. If hospitals inside the frame were similar to hospitals outside the frame, the sample hospitals can be treated as if they were randomly selected from the entire universe of hospitals within each stratum. Discharges were randomly selected from within each hospital. Standard formulas for stratified, two-stage cluster sample without replacement may be used to calculate statistics and their variances in most applications.
Examples of the use of SAS, SUDAAN, and STATA to calculate variances in the KID are presented in the special report: Calculating Kids’ Inpatient Database Variances. This report is available on the KID Documentation CD-ROM and on the HCUP User Support Website at http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp.
Studying Trends
When studying trends over time using the KID, be aware that the sampling frame for the KID changes over time. Because more states have been added, estimates from earlier years of the KID may be subject to more sampling bias than later years of the KID.
Short-term rehabilitation hospitals are included in the 1997 KID, but are excluded from the 2000 and 2003 KID. Patients treated in short-term rehabilitation hospitals tend to have lower mortality rates and longer lengths of stay than patients in other community hospitals. The elimination of rehabilitation hospitals may affect trends but the effect is likely small since only about three percent of community hospitals are short-term rehabilitation hospitals and not all state data sources included short term rehabilitation hospitals.
The Kids’ Inpatient Database (KID) is one of a family of databases and software tools developed as part of the Healthcare Cost and Utilization Project (HCUP), a Federal-State-Industry partnership sponsored by the Agency for Healthcare Research and Quality (AHRQ).
The KID is a unique and powerful nationwide database of hospital inpatient stays for children. It is a sample of pediatric discharges (age 20 or less at admission) from community, non-rehabilitation hospitals from states participating in HCUP. The HCUP KID team developed the database to permit researchers to study a broad range of conditions and procedures related to hospitalizations of children. Researchers and policy makers can use the KID to identify, track, and analyze national trends in hospital utilization, access, charges, quality, and outcomes for children.
Potential research issues focus on both discharge- and hospital-level outcomes. Discharge outcomes of interest include:
Hospital outcomes of interest include:
These and other outcomes are of interest for the nation as a whole and for policy-relevant inpatient subgroups defined by diagnoses and procedures, geographic region, patient demographics, hospital characteristics, and pay sources.
This report focuses on the KID sample and weights, summarizes the contents of the 2003 KID, and discusses data analysis issues. Previous KID releases covered 1997 and 2000. We have enhanced the nationwide representation of the sample by incorporating data from additional HCUP State Partners. The 2003 KID includes data from 36 states, nine more than the 2000 KID. This document highlights cumulative information for all previous years to provide a longitudinal view of the database.
The hospital universe is defined as all hospitals located in the U.S. that were open during any part of the calendar year and that were designated as community hospitals in the American Hospital Association (AHA) Annual Survey. The AHA defines community hospitals as follows: "All nonfederal short-term general and other specialty hospitals, excluding hospital units of institutions." Consequently, Veterans Hospitals and other Federal facilities (Department of Defense and Indian Health Service) are excluded. Beginning with the 2000 KID, short-term rehabilitation hospitals were excluded from the universe, because the type of care provided and the characteristics of the discharges from these facilities were markedly different from other short-term hospitals. (The 1997 KID includes short-term rehabilitation hospitals.) Table 1 displays the number of hospitals in the universe for each year, based on the corresponding AHA Annual Survey.
| Year | Number of Hospitals |
|---|---|
| 1997 | 5,113 |
| 2000 | 4,839 |
| 2003 | 4,836 |
All U.S. hospital entities that were designated community hospitals in the AHA hospital file, except short-term rehabilitation hospitals, were included in the hospital universe for the 2003 KID. Therefore, when two or more hospitals merged to create a new hospital, the original hospitals and the newly-formed hospital were all considered separate entities in the universe during the year they merged. Similarly, if a hospital split, the original hospital and all newly-created hospitals were separate entities in the universe during the year they split. Finally, hospitals that closed during a year were included as long as they were in operation during some part of the calendar year.
For the purpose of calculating discharge weights, we post-stratified hospitals on six characteristics contained in the AHA hospital files. These were the same characteristics used to define the HCUP Nationwide Inpatient Sample (NIS) sampling strata, with the addition of a stratum for stand-alone children’s hospitals. The definitions of some of the NIS strata were revised for 1998 and subsequent data years, and we used the revised strata beginning with the 2000 KID. (A description of the strata used for the 1997 KID can be found in the Kids Inpatient Database (KID) Design Report, 1997. This report is available on the 1997 KID Documentation CD-ROM and on the HCUP Website at http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp.)
Beginning with the 2000 KID, the stratification variables were defined as follows:
Figure 1: KID States, by Region (text version)
| Region | States |
|---|---|
| 1: Northeast | Connecticut , Maine, Massachusetts, New Hampshire, New Jersey, New York, Pennsylvania, Rhode Island, Vermont |
| 2: Midwest | Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, Wisconsin |
| 3: South | Alabama, Arkansas, Delaware, District of Columbia, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, West Virginia |
| 4: West | Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, Wyoming |
| Location and Teaching Status | Hospital Bed Size | ||
|---|---|---|---|
| Small | Medium | Large | |
| NORTHEAST | |||
| Rural | 1-49 | 50-99 | 100+ |
| Urban, non-teaching | 1-124 | 125-199 | 200+ |
| Urban, teaching | 1-249 | 250-424 | 425+ |
| MIDWEST | |||
| Rural | 1-29 | 30-49 | 50+ |
| Urban, non-teaching | 1-74 | 75-174 | 175+ |
| Urban, teaching | 1-249 | 250-374 | 375+ |
| SOUTH | |||
| Rural | 1-39 | 40-74 | 75+ |
| Urban, non-teaching | 1-99 | 100-199 | 200+ |
| Urban, teaching | 1-249 | 250-449 | 450+ |
| WEST | |||
| Rural | 1-24 | 25-44 | 45+ |
| Urban, non-teaching | 1-99 | 100-174 | 175+ |
| Urban, teaching | 1-199 | 200-324 | 325+ |
The universe of hospitals was established as all community hospitals located in the U.S. with the exception, beginning in 2000, of short-term rehabilitation hospitals. However, some hospitals do not supply data to HCUP. Therefore, we constructed the KID sampling frame from the subset of universe hospitals that released their discharge data to AHRQ for research use. AHRQ obtained agreements with 36 HCUP State Partner organizations to include their data in the 2003 KID. The number of State Partners and hospitals contributing data to the KID has expanded over the years, as shown in Table 4.
| Calendar Year | States in the Frame | Number of States | Sample Hospitals | Sample Discharges |
|---|---|---|---|---|
| 1997 | Arizona, California, Colorado, Connecticut, Florida, Georgia, Hawaii, Illinois, Iowa, Kansas, Maryland, Massachusetts, Missouri, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Tennessee, Utah, Washington, and Wisconsin. | 22 | 2,521 | 1,905,797 |
| 2000 | Added Kentucky, Maine, North Carolina, Texas, Virginia, and West Virginia. Dropped Illinois. | 27 | 2,784 | 2,516,833 |
| 2003 | Added Illinois, Indiana, Michigan, Minnesota, Nebraska, New Hampshire, Nevada, Ohio, Rhode Island, South Dakota, and Vermont. Dropped Maine and Pennsylvania. | 36 | 3,438 | 2,984,129 |
The list of the entire frame of hospitals was composed of all AHA community, non-rehabilitation hospitals in each of the frame states that could be matched to the discharge data provided to HCUP. If an AHA hospital could not be matched to the discharge data provided by the data source, it was eliminated from the sampling frame (but not from the target universe).
Table 5 shows the number of AHA, HCUP SID, and KID hospitals by state. The columns in Table 5 are defined are as follows:
In most cases, the difference between the universe and the frame represents the difference between the number of community, non-rehabilitation hospitals in the 2003 AHA Annual Survey of Hospitals and the number of hospitals with children’s discharges that were supplied to HCUP that could be matched to the AHA data.
The largest discrepancy between HCUP data and AHA data is in Texas. As is evident in Table 5, only 303 out of 414 Texas community, non-rehabilitation hospitals supplied data to HCUP for 2003. Certain Texas state-licensed hospitals are exempt from statutory reporting requirements. Exempt hospitals include:
The Texas statute that exempts rural providers from the requirement to submit data defines a hospital as a rural provider if it:
These exemptions apply primarily to smaller rural public hospitals and, as a result, these facilities are less likely to be included in the sampling frame than other Texas hospitals. While the number of hospitals omitted appears sizable, those available for the KID include 91.8% of inpatient discharges from Texas universe hospitals.
The Minnesota frame contains 15 fewer hospitals than the state universe because a few of the state’s hospitals do not participate in HCUP. There are no apparent significant differences between the characteristics of participating and non-participating Minnesota hospitals.
The Ohio frame contains 11 fewer hospitals than the state universe, including three hospitals that could not be matched to the AHA data because the Partner masked their identities in the data.
For Connecticut, Georgia, Hawaii, Indiana, Michigan, Nebraska, South Carolina, South Dakota, and Virginia, we had to exclude several HCUP hospitals from the frame, as described below.
To help ensure the confidentiality of hospitals in Connecticut, the Partner requested that we exclude the only stand-alone community children’s hospital in the state from the database.
Georgia, Hawaii, Indiana, Michigan, Nebraska, South Carolina, and South Dakota stipulated that only hospitals appearing in sampling strata with two or more hospitals from the state were to be included in the KID. Because of this restriction, two Georgia hospitals (including the only freestanding community, non-rehabilitation children’s hospital in the state), four Hawaii hospitals, one Indiana hospital, one Michigan hospital, three Nebraska hospitals, five South Carolina hospitals, and two South Dakota hospitals were excluded from the KID sampling frame. An additional 43 Georgia hospitals were randomly omitted from the KID sampling frame because Georgia allowed no more than 60% of the state’s hospitals to be included in the KID. The two stand-alone children’s hospitals in Nebraska were also excluded at the Partner’s request. Two additional South Carolina hospitals were removed from the sampling frame because they have unique characteristics that might make them identifiable.
Because charges represent a critical outcome variable in the KID, we decided to omit 32 additional Michigan hospitals from the frame that did not provide total charges. By excluding these hospitals, we avoid having to adjust the weights or create another weighting variable specifically for total charges. These hospitals are fairly evenly distributed by hospital type. There are no sampling strata in the state containing only hospitals without charges. The total charge data reported for Michigan is similar to total charge data reported by other Midwestern states. Thus, there does not seem to be an obvious bias in the type of cases for which charges are reported. The stratification and weighting scheme should adjust for the hospitals that are being excluded.
Because Virginia allowed only 50% or fewer of the state’s hospitals to be included in the KID, 35 hospitals from the state were randomly omitted from the KID sampling frame.
A total of 133 hospitals were restricted from the KID sampling frame, leaving 3,445 hospitals with pediatric discharges in the frame. Seven of these hospitals were not represented in the KID because they had so few pediatric discharges that none were sampled.
Beginning with the 2000 KID, pediatric discharges were defined as having an age at admission of 20 or less. This differs from the 1997 KID, which included discharges with an admission age of 18 or less. Discharges with missing, invalid, or inconsistent ages were excluded.
| State | AHA Universe Hospitals | SID Community, Non-Rehabilitation Hospitals | SID Community, Non-Rehabilitation Hospitals with Pediatric Discharges | KID Sampling- Frame Hospitals | KID Sample Hospitals |
|---|---|---|---|---|---|
| Non-Frame | 987 | 215 | 215 | 0 | 0 |
| Arizona | 59 | 58 | 58 | 58 | 58 |
| California | 367 | 361 | 359 | 359 | 358 |
| Colorado | 66 | 65 | 64 | 64 | 64 |
| Connecticut | 34 | 31 | 31 | 30 | 30 |
| Florida | 194 | 191 | 188 | 188 | 188 |
| Georgia | 143 | 142 | 141 | 96 | 96 |
| Hawaii | 23 | 20 | 20 | 16 | 16 |
| Iowa | 116 | 116 | 113 | 113 | 113 |
| Illinois | 189 | 188 | 187 | 187 | 187 |
| Indiana | 110 | 107 | 106 | 105 | 105 |
| Kansas | 136 | 128 | 128 | 128 | 127 |
| Kentucky | 98 | 96 | 96 | 96 | 96 |
| Massachusetts | 73 | 67 | 66 | 66 | 66 |
| Maryland | 48 | 46 | 46 | 46 | 46 |
| Michigan | 142 | 132 | 129 | 96 | 95 |
| Minnesota | 131 | 116 | 115 | 115 | 114 |
| Missouri | 118 | 115 | 114 | 114 | 114 |
| North Carolina | 113 | 109 | 109 | 109 | 109 |
| Nebraska | 84 | 83 | 82 | 77 | 77 |
| New Hampshire | 26 | 26 | 26 | 26 | 26 |
| New Jersey | 73 | 73 | 73 | 73 | 73 |
| Nevada | 26 | 25 | 25 | 25 | 25 |
| New York | 204 | 203 | 200 | 200 | 200 |
| Ohio | 161 | 150 | 150 | 150 | 150 |
| Oregon | 58 | 56 | 56 | 56 | 56 |
| Rhode Island | 11 | 11 | 11 | 11 | 11 |
| South Carolina | 59 | 57 | 57 | 50 | 50 |
| South Dakota | 50 | 47 | 45 | 43 | 43 |
| Tennessee | 121 | 116 | 114 | 114 | 114 |
| Texas | 414 | 303 | 275 | 275 | 273 |
| Utah | 41 | 40 | 40 | 40 | 40 |
| Virginia | 81 | 79 | 79 | 44 | 44 |
| Vermont | 14 | 14 | 14 | 14 | 14 |
| Washington | 85 | 84 | 84 | 84 | 83 |
| Wisconsin | 127 | 127 | 125 | 125 | 125 |
| West Virginia | 54 | 54 | 52 | 52 | 52 |
| Total | 4,836 | 3,851 | 3,793 | 3,445 | 3,438 |
The overall design objective was to select a sample of pediatric discharges that accurately represents the target universe of U.S. community, non-rehabilitation hospitals. Moreover, this sample was to be geographically dispersed, yet drawn exclusively from hospitals in states that participate in HCUP and agree to contribute to the KID.
It should be possible, for example, to estimate DRG-specific average lengths of stay across all U.S. hospitals using weighted average lengths of stay, based on averages or regression coefficients calculated from the KID. Ideally, relationships among outcomes and their correlates estimated from the KID should accurately represent all U.S. hospitals. However, the 2003 KID includes data from only 36 states. Therefore, it is advisable to verify your estimates against other data sources, if available. For example, the National Hospital Discharge Survey (http://www.cdc.gov/nchs/about/major/hdasd/nhds.htm) can provide benchmarks against which to check your national estimates for hospitalizations with more than 5,000 cases. The KID Comparison Report assesses the accuracy of KID estimates. The most recent report is available on the KID Documentation CD-ROM and provides a comparison of the 1997 KID with other data sources. The updated report for the current KID is expected to be available on the HCUP User Support Website (http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp) in late January, 2006.
In order to sample and project births up to the number of births reported by the AHA, which reports in-hospital births, the KID development team identified all in-hospital births in the KID data. We further separated the in-hospital births in HCUP data into uncomplicated births and complicated births. We sampled uncomplicated births at a lower rate because they have little variation in their outcomes.
To determine the best way to identify in-hospital births, we ran cross-tabulations of different combinations of variables on all cases that had any of the following possible birth indicators: age of zero days (AGEDAY=0), neonatal diagnosis (NEOMAT>=2), neonatal Major Diagnostic Category (MDC 15), or admission type of birth (ATYPE=43). Based on reviews of the cross-tabulations, the MDC 15 DRG definitions, and ICD-9-CM birth diagnosis codes, the following screen was devised for births: an in-hospital birth diagnosis code (any diagnosis code in the range V3000 - V3901 with a fourth digit of zero, indicating born in the hospital, and a fifth digit of zero or one, indicating delivered without mention of cesarean delivery, or delivered by cesarean delivery), without an admission source of another hospital or health facility (ASOURCE not equal to 2 or 3).
We classified neonates transferred from other facilities as pediatric non-births because they are not included in births reported by the AHA. An age of zero days was not a reliable in-hospital birth indicator because neonates transferred from another hospital or born before admission to the hospital could also have an age of zero days. There were also some cases with birth diagnoses, but with ages of a few days. Because the HCUP data are already edited for neonatal diagnoses inconsistent with age, we did not include any age criteria in the in-hospital birth screen.
Uncomplicated, in-hospital births are identified as cases that meet the above screen and are in DRG 391, "Normal Newborn." In the 2003 KID, approximately 1.2% of the cases in DRG 391 do not meet the in-hospital birth screen. These cases have diagnoses that imply a newborn, but do not specifically indicate an in-hospital birth. It is possible that some of these may have actually been born in the hospital but lacked the proper diagnosis code. Others may be readmissions or may have been born before admission to the hospital. Less than 0.2% of cases in DRG 391 have an admission type of newborn (ATYPE = 4) but do not meet the in-hospital birth screen.
Using the above in-hospital birth screen, we identified 3,249,450 in-hospital births in community, non-rehabilitation hospitals in the 2003 KID, as compared with 3,238,145 births reported by the AHA in these hospitals. There were 11,305 fewer births reported by the AHA, a difference of less than .4%.
We use the NIS community hospital universe and strata definitions for the KID. We revised some of the NIS hospital universe and strata definitions for 1998 and subsequent data years, and we used these revised definitions beginning with the 2000 KID. These changes included:
A full description of the evaluation and revision of the NIS sampling strategy for 1998 and subsequent data years can be found in the special report, Changes in NIS Sampling and Weighting Strategy for 1998. This document is available on the 2003 KID Documentation CD-ROM and on the HCUP User Support Website at http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp.
The KID includes a sample of pediatric discharges from all hospitals in the sampling frame. For the sampling, we stratified the pediatric discharges by uncomplicated in-hospital birth, complicated in-hospital birth, and pediatric non-birth. To further ensure an accurate representation of each hospital’s pediatric case-mix, we also sorted the discharges by state, hospital, DRG, and a random number within each DRG. We then used systematic random sampling to select 10 percent of uncomplicated in-hospital births and 80 percent of other pediatric cases from each frame hospital.
It should be observed that the NIS includes 100% of the discharges from hospitals in the NIS sample. Consequently, in the NIS outcomes can be estimated without sampling error for individual hospitals that are identified in the sample. However, the KID includes fewer than 100% of the pediatric discharges for each hospital in the database. Therefore, researchers will not be able to calculate hospital-specific outcomes with certainty.
To obtain national estimates, we developed discharge weights using the AHA universe as the standard. For the weights, we post-stratified hospitals on six characteristics contained in the AHA hospital files. These were the same characteristics used to define the NIS sampling strata, with the addition of a stratum for freestanding children’s hospitals. We also stratified the KID discharges according to whether the discharge was an uncomplicated in-hospital birth, a complicated in-hospital birth, or a non-newborn pediatric discharge. If there were fewer than two frame hospitals, 30 uncomplicated births, 30 complicated births, and 30 non-birth pediatric discharges sampled in a stratum, we merged that stratum with an "adjacent" stratum containing hospitals with similar characteristics.
The discharge weights were created by stratum, in proportion to the number of AHA discharges for newborns and non-newborns. Refer to the report Design of the HCUP Kids’ Inpatient Database (KID), 1997, for a discussion of the analysis and development of the KID weighting scheme. This report is available on the 1997 KID Documentation CD-ROM and on the HCUP Website at http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp.
We used NACHRI data to help verify and correct the AHA list of children’s hospitals in the target universe. Many of these children’s hospitals are units of larger institutions (AHA hospital type 10). Consequently, we do not have separate reporting for them either in the AHA survey or in the HCUP SID. However, data analysts may find it useful to identify hospitals that contain children’s units, which can be accomplished using the NACHTYPE variable in the KID.
The discharge weights usually are constant for all discharges of the same type (uncomplicated in-hospital birth, complicated in-hospital birth, and other pediatric discharge) within a stratum. The only exceptions are for strata with sample hospitals that, according to the AHA files, were open for the entire year but contributed less than their full year of data to the KID. For those hospitals, we adjusted the number of observed discharges by a factor of 4 ÷ Q, where Q was the number of calendar quarters that the hospital contributed discharges to the KID. For example, when a sample hospital contributed only two quarters of discharge data to the KID, the adjusted number of discharges was double the observed number.
With that minor adjustment, each discharge weight is essentially equal to the number of AHA universe discharges that each sampled discharge represents in its stratum. This calculation was possible because the numbers of total discharges and births were available for every hospital in the universe from the AHA files.
Discharge weights to the universe were calculated by post-stratification. Hospitals were stratified on geographic region, urban/rural location, teaching status, bed size, control, and hospital type. In some instances, strata were collapsed for sample weight calculations. Within stratum k, for hospital i, each KID sample discharge's universe weight was calculated as:
Wik = [Tk / (Rk * Ak)] * (4 ÷ Qi)
In the birth strata (both complicated and uncomplicated):
In the non-newborn strata:
Qi is the number of quarters of discharge data contributed by hospital i to the KID (usually Qi = 4).
Tk / Ak estimates the number of discharges in the population that is represented by each discharge in the sampling frame. Rk adjusts for the fact that we are taking a sample of the frame in each stratum.
Uncomplicated in-hospital births were sampled at a lower rate than other discharges because the variation in hospital outcomes for uncomplicated births is considerably less than that for other pediatric cases and because we expect research to focus much more on other pediatric patients. We sampled uncomplicated births at the nominal rate of 10 percent and sampled other pediatric discharges (complicated newborns and other pediatric cases) at the nominal rate of 80 percent from the discharges available in the (restricted) frame. To avoid rounding errors in the weights calculation, the actual sampling rate for a discharge type (uncomplicated in-hospital birth, complicated in-hospital birth, or non-birth pediatric discharge) in stratum k, Rk, was calculated as follows:
Rk = Sk / Hk
The AHA birth counts include both uncomplicated and complicated births. Therefore, the weights in the uncomplicated birth strata implicitly assume that the proportion of births that are uncomplicated in the frame is representative of the proportion of births that are uncomplicated in the population for each stratum. A similar assumption is made for complicated newborns.
Similarly, the non-birth AHA discharge counts include all non-birth discharges, not just non-birth pediatric discharges. Consequently, the weights in the non-birth strata implicitly assume that the proportion of non-birth discharges that are pediatric across the HCUP SID hospitals is the same as the proportion of non-birth discharges that are pediatric across the universe of AHA hospitals, in the aggregate within each hospital stratum.
To produce nationwide estimates, use the discharge weights to extrapolate sampled discharges in the Core file to the discharges from all U.S. community, non-rehabilitation hospitals. For the 2003 KID, use DISCWT to calculate nationwide estimates for all analyses. For the 2000 KID, use DISCWT to create nationwide estimates for all analyses except those that involve total charges, and use DISCWTCHARGE to create nationwide estimates of total charges. For the 1997 KID, use DISCWT_U for all analyses.
There were 3,445 hospitals in the 2003 sampling frame, a 23.6% increase from the 2000 KID. The final 2003 KID sample included 2,984,129 discharges of children from 3,438 hospitals drawn from 36 frame states representing all regions of the United States. The 2003 KID is larger than the 2000 KID across several dimensions:
Table 6 summarizes the numbers of hospitals and discharges for children’s hospitals and other hospitals. For each hospital type, the table shows the number of:
| Hospital Type | AHA Universe | SID | KID | |||
|---|---|---|---|---|---|---|
| Hospitals | Discharges (Including Births) | Hospitals with Pediatric Discharges | Pediatric Discharges | Hospitals | Pediatric Discharges | |
| Not a Children's Hospital | 4,757 | 37,689,130 | 3,742 | 6,034,209 | 3,396 | 2,714,359 |
| Children's Hospital | 79 | 531,529 | 51 | 422,072 | 42 | 269,770 |
| Total | 4,836 | 38,220,659 | 3,793 | 6,456,281 | 3,438 | 2,984,129 |
Figure 2 summarizes the 2003 KID hospitals by geographic region. For each region, the chart presents:
Although pediatric discharges from hospitals in each region are selected for the KID, the comprehensiveness of the sampling frame varies by region. Figure 2 reveals that the percentage of hospitals included in the sampling frame is highest in the Midwest (89.0%) and in the West (78.0%), while figures are lower for the Northeast (63.9%) and the South (56.9%).
Because the KID sampling frame has a disproportionate representation of the more populous states and contains hospitals with more annual discharges, its comprehensiveness in terms of discharges is higher. Figure 3 summarizes the estimated U.S. population by geographic region on July 1, 20034. For each region, the figure displays:
For example, the estimated population of the Northeast region on July 1, 2003 was 54,426,252. On that same date, the estimated population of states in the Northeast region that were included in the 2003 KID was 40,746,286. This represents 74.9% of the total Northeast region’s population. The percentage of the Northeast population represented declined from 94.6% in the 2000 KID because the addition of New Hampshire, Rhode Island, and Vermont was offset by the loss of Maine and Pennsylvania. However, the seven Midwestern states added for 2003 have substantially increased the percentage of the Midwestern population represented, from 25.7% in the 2000 KID to 99.0% in the 2003 KID. The percentage of estimated U.S. population included in the West (92.0%) and in the South (81.3%) was slightly higher than for 2000. Overall, the states in the 2003 KID include an estimated 86.5% of the entire U.S population, up from 76.2% in the 2000 KID.
Figure 4 presents the number of discharges in the KID for each state in the sampling frame for 2003. The number of discharges ranges from 4,724 in Vermont to 428,650 in California.
Figure 2: Number of Hospitals in the 2003 AHA Universe, SID, and KID, by Region (text version)
Figure 3: Percentage of U.S. Population in 2003 KID States, by Region (text version)
Figure 4: Number of Discharges in the 2003 KID, by State (text version)
Table 7 displays the unweighted and weighted number of uncomplicated births, complicated births, and pediatric non-births by hospital type in the 2003 KID.
| Hospital Type | Uncomplicated Births | Complicated Births | Pediatric Non-Births | Total Pediatric Discharges |
|---|---|---|---|---|
| Unweighted: | ||||
| Not a Children's Hospital | 243,494 | 645,064 | 1,825,801 | 2,714,359 |
| Children's Hospital | 344 | 1,680 | 267,746 | 269,770 |
| Total | 243,838 | 646,744 | 2,093,547 | 2,984,129 |
| Weighted: | ||||
| Not a Children's Hospital | 2,937,048 | 979,523 | 2,976,017 | 6,892,588 |
| Children's Hospital | 2,616 | 1,600 | 512,358 | 516,574 |
| Total | 2,939,665 | 981,122 | 3,488,375 | 7,409,162 |
Missing data values can compromise the quality of estimates. If the outcome for discharges with missing values is different from the outcome for discharges with valid values, then sample estimates for that outcome will be biased and inaccurately represent the discharge population. There are several techniques available to help overcome this bias. One strategy is to use imputation to replace missing values with acceptable values. Another strategy is to use sample weight adjustments to compensate for missing values. Descriptions of such data preparation and adjustment are outside the scope of this report; however, it is recommended that researchers evaluate and adjust for missing data, if necessary.
On the other hand, if the cases with and without missing values are assumed to be similar with respect to their outcomes, no adjustment may be necessary for estimates of means and rates. This is because the non-missing cases would be representative of the missing cases. However, some adjustment may still be necessary for the estimates of totals. Sums of data elements (such as aggregate charges) containing missing values would be incomplete because cases with missing values would be omitted from the calculations.
It may be important for researchers to calculate a measure of precision for some estimates based on the KID sample data. Variance estimates must take into account both the sampling design and the form of the statistic. If hospitals inside the frame are similar to hospitals outside the frame, the sample hospitals can be treated as if they were randomly selected from the entire universe of hospitals within each stratum. Discharges were randomly selected from within each hospital. Standard formulas for stratified, two-stage cluster samples without replacement may be used to calculate statistics and their variances in most applications.
A multitude of statistics can be estimated from the KID data. Several computer programs that calculate statistics and their variances from sample survey data are listed in the section below. Some of these programs use general methods of variance calculations (e.g., the jackknife and balanced half-sample replications) that take into account the sampling design. However, it may be desirable to calculate variances using formulas specifically developed for some statistics.
These variance calculations are based on finite-sample theory, which is an appropriate method for obtaining cross-sectional, nationwide estimates of outcomes. According to finite-sample theory, the intent of the estimation process is to obtain estimates that are precise representations of the nationwide population at a specific point in time. In the context of the KID, any estimates that attempt to accurately describe characteristics (such as expenditure and utilization patterns or hospital market factors) and interrelationships among characteristics of hospitals and discharges during a specific year should be governed by finite-sample theory.
Alternatively, in the study of hypothetical population outcomes not limited to a specific point in time, the concept of a "superpopulation" may be useful. Analysts may be less interested in specific characteristics from the finite population (and time period) from which the sample was drawn than they are in hypothetical characteristics of a conceptual superpopulation from which any particular finite population in a given year might have been drawn. According to this superpopulation model, the nationwide population in a given year is only a snapshot in time of the possible interrelationships among hospital, market, and discharge characteristics. In a given year, all possible interactions between such characteristics may not have been observed, but analysts may wish to predict or simulate interrelationships that may occur in the future.
Under the finite-population model, the variances of estimates approach zero as the sampling fraction approaches one. This is the case because the population is defined at that point in time, and because the estimate is for a characteristic as it existed when sampled. This is in contrast to the superpopulation model, which adopts a stochastic viewpoint rather than a deterministic viewpoint. That is, the nationwide population in a particular year is viewed as a random sample of some underlying superpopulation over time. Different methods are used for calculating variances under the two sample theories. The choice of an appropriate method for calculating variances for nationwide estimates depends on the type of measure and the intent of the estimation process.
The discharge weights would be used to weight the sample data in estimating population statistics. In most cases, computer programs are readily available to perform these calculations. Several statistical programming packages allow weighted analyses. For example, nearly all Statistical Analysis System (SAS) procedures incorporate weights. In addition, several statistical analysis programs have been developed to specifically calculate statistics and their standard errors from survey data. Version eight or later of SAS contains procedures (PROC SURVEYMEANS and PROC SURVEYREG) for calculating statistics based on specific sampling designs. STATA and SUDAAN are two other common statistical software packages that perform calculations for numerous statistics arising from the stratified, single-stage cluster sampling design. Examples of the use of SAS, SUDAAN, and STATA to calculate KID variances are presented in the special report: Calculating Kids’ Inpatient Database Variances. This report is available on the 2003 KID Documentation CD-ROM and on the HCUP User Support Website at http://www.hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp. For an excellent review of programs to calculate statistics from survey data, visit the following Website: http://www.hcp.med.harvard.edu/statistics/survey-soft/. 
The KID database includes a Hospital file with variables required to calculate finite population statistics. The file includes hospital identifiers (Primary Sampling Units or PSUs), stratification variables, and stratum-specific totals for the numbers of discharges and hospitals so that finite-population corrections can be applied to variance estimates.
In addition to these subroutines, standard errors can be estimated by validation and cross-validation techniques. Given that a very large number of observations will be available for most analyses, it may be feasible to set aside a part of the data for validation purposes. Standard errors and confidence intervals can then be calculated from the validation data.
If the analytical file is too small to set aside a large validation sample, cross-validation techniques may be used. For example, tenfold cross-validation would split the data into ten equal-sized subsets. The estimation would take place in ten iterations. In each iteration, the outcome of interest is predicted for one-tenth of the observations by an estimate based on a model fit to the other nine-tenths of the observations. Unbiased estimates of error variance are then obtained by comparing the actual values to the predicted values obtained in this manner.
Finally, it should be noted that a large array of hospital-level variables are available for the entire universe of hospitals, including those outside the sampling frame. For instance, the variables from the AHA surveys and from the Medicare Cost Reports are available for nearly all hospitals in the U.S, although hospital identifiers are suppressed in the KID for a number of states. For these states it will not be possible to link to outside hospital-level data sources. To the extent that hospital-level outcomes correlate with these variables, they may be used to sharpen regional and nationwide estimates.
When studying trends over time using the KID, be aware that the sampling frame for the KID changes over time. Because more states have been added, estimates from earlier years of the KID may be subject to more sampling bias than later years of the KID.
Short-term rehabilitation hospitals are included in the 1997 KID, but are excluded from the 2000 and 2003 KID. Patients treated in short-term rehabilitation hospitals tend to have lower mortality rates and longer lengths of stay than patients in other community hospitals. The elimination of rehabilitation hospitals may affect trends but the effect is likely small since only about three percent of community hospitals are short-term rehabilitation hospitals and not all state data sources included short term rehabilitation hospitals.
In this report, we have described the development and use of the KID sample and weights and summarized the contents of the 2003 KID. We have included cumulative information for all previous years to provide a longitudinal view of the database. Once again, the nationwide representation of the database has been enhanced by incorporating data from additional HCUP State Partners, a total of 36 participants for the year 2003. We have highlighted important considerations for data analysis and have provided references to detailed reports on this subject.
Email Updates
/* FIRST ESTABLISH SHORT-TERM
BEDS DEFINITION */
IF BDH NE
. THEN BEDTEMP = BDH ; /* SHORT TERM BEDS */
ELSE IF BDH =. THEN BEDTEMP=BDTOT ; /* TOTAL BEDS PROXY
*/
/*******************************************************/
/* NEXT ESTABLISH
TEACHING STATUS BASED ON F-T & P-T */
/* RESIDENT/INTERN
STATUS FOR HOSPITALS. */
/*******************************************************/
RESINT = (FTRES + .5*PTRES)/BEDTEMP
;
IF (MAPP3 = . AND MAPP8 =
.) THEN DO ;
IF RESINT > .10 THEN ST_TEACH = 1 ;
ELSE ST_TEACH = 0 ;
END ;
IF (MAPP3=1 OR MAPP8=1) THEN
ST_TEACH=1 ; /* 1=TEACHING */
ELSE ST_TEACH=0 ; /* 0=NONTEACHING */
/*******************************************************/
/* CREATE TEACHING CATEGORY
VARIABLES TO FURTHER */
/* REFINE TEACHING STATUS
DEFINITION. */
/*******************************************************/
IF ST_TEACH = 1 THEN DO ;
IF
0 < RESINT < .15 THEN TEACHCAT=0 ; /* MINOR CATEGORY */
ELSE
IF RESINT GE .15 THEN TEACHCAT=1 ; /* MAJOR CATEGORY */
ELSE
ST_TEACH = 0 ; /* NONTEACH STATUS*/
END ;
/*******************************************************/
/* FIRST ESTABLISH SHORT-TERM BEDS DEFINITION */
/*******************************************************/
IF BDH NE . THEN BEDTEMP = BDH ; /* SHORT TERM BEDS */
ELSE IF BDH =. THEN BEDTEMP = BDTOT ; /* TOTAL BEDS PROXY */
/*******************************************************/
/* ESTABLISH IRB NEEDED FOR TEACHING STATUS */
/* BASED ON F-T P-T RESIDENT INTERN STATUS */
/*******************************************************/
IRB = (FTRES + .5*PTRES) / BEDTEMP ;
/*******************************************************/
/* CREATE TEACHING STATUS VARIABLE */
/*******************************************************/
IF (MAPP8 EQ 1) OR (MAPP3
EQ 1) THEN HOSP_TEACH = 1 ;
ELSE IF (IRB GE 0.25) THEN
HOSP_TEACH = 1 ;
ELSE HOSP_TEACH = 0;
| Internet Citation: 2003 KID Design Report. Healthcare Cost and Utilization Project (HCUP). June 2016. Agency for Healthcare Research and Quality, Rockville, MD. www.hcup-us.ahrq.gov/db/nation/kid/reports/KID_design_rpt_2003.jsp. |
| Are you having problems viewing or printing pages on this website? |
| If you have comments, suggestions, and/or questions, please contact hcup@ahrq.gov. |
| Privacy Notice, Viewers & Players |
| Last modified 6/23/16 |