PEDIATRICS Vol. 121 No. 6 June 2008, pp. e1653-e1659 (doi:10.1542/peds.2007-2831)
ARTICLE |
Charges and Lengths of Stay Attributable to Adverse Patient-Care Events Using Pediatric-Specific Quality Indicators: A Multicenter Study of Freestanding Children's Hospitals
a Divisions of General Pediatrics and Infectious Diseases, Children's Hospital of Philadelphia, Philadelphia, Pennsylvania
b Child Health Corporation of America, Shawnee Mission, Kansas
c Departments of Internal Medicine and Pediatrics, Carilion Clinic Children's Hospital, Roanoke, Virginia
d Departments of Pediatrics and Epidemiology and Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
OBJECTIVE. The purpose of this work was to determine the excess charges, both overall and according to category, and lengths of stay attributable to adverse patient-care events during pediatric hospitalization.
METHODS. Agency for Healthcare Research and Quality pediatric-specific quality indicators were used to identify adverse events in 431524 discharges from 38 freestanding, academic, not-for-profit, tertiary care pediatric hospitals in the United States participating in the Pediatric Health Information System database in 2006. All of the discharges from any of the 38 hospitals participating in the Pediatric Health Information System between January 1 and December 31, 2006, were eligible for inclusion. The primary outcomes were excess lengths of stay and charges (both overall and according to pharmacy, laboratory, imaging, clinical, supply, and other categories) were attributable to adverse patient-safety events as determined by 12 pediatric-specific quality indicators.
RESULTS. Statistically significant excess lengths of stay attributable to pediatric-specific quality indicator events ranged from 2.8 days for accidental puncture and laceration to 23.5 days for postoperative sepsis, and statistically significant excess overall charges ranged from $34884 for accidental puncture and laceration to $337226 for in-hospital mortality after pediatric heart surgery. Each charge category had significant charge increases caused by pediatric-specific quality indicator events, with the largest being laboratory and other charges, ranging from $7622 to $78048 and $11094 to $97805, respectively.
CONCLUSIONS. Some adverse events experienced during pediatric hospitalization have the potential to increase lengths of stay and charges considerably, and pediatric-specific quality indicators are useful in calculating these effects.
Key Words: patient-safety indicators • quality indicators • safety • medical error • inpatients • pediatrics • children • administrative data
Abbreviations: AHRQ—Agency for Healthcare Research and Quality • PSI—patient-safety indicator • ICD-9-CM—International Classification of Diseases, Ninth Revision, Clinical Modification • LOS—length of stay • PDI—pediatric-specific quality indicator • PHIS—Pediatric Health Information System • APR-DRG—all-patient refined diagnosis-related group • KID—Kids' Inpatient Database
In 2003, the Agency for Healthcare Research and Quality (AHRQ) released a set of patient-safety indicators (PSIs), a tool using administrative data to identify potentially adverse patient-care events.1,2 These indicators use International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnoses and procedure codes recorded as part of the administrative data. The PSIs were specifically developed to target events likely to represent preventable medical errors, such as an iatrogenic pneumothorax or a foreign body left in during a procedure. The PSIs were initially tested and validated in adults.2 The goal of these PSIs was to allow institutions to track and evaluate such events to improve patient safety; the AHRQ has revised the PSIs over time to increase their use.
Several adult studies used PSIs to determine rates of AHRQ-defined quality events among Veteran's Administration patients3,4 and to demonstrate differences in events across racial and ethnic lines.5 Miller and Zhan6 used PSIs to ascertain the impact that adverse patient-safety events had on length of stay (LOS), charges, and mortality during hospitalizations for a large and diverse group of patients. They subsequently reported rates of PSI events in a pediatric patient population,7 and, along with Elixhauser,8 reported the excess charges, LOS, and mortality associated with these pediatric events. Sedman et al9 later demonstrated, however, that the original AHRQ PSIs might be inappropriate for the pediatric population.
Recognizing that PSIs derived for adults were not always applicable to children, the AHRQ revised the PSIs developed for the adult population in September 2006 to create a pediatric-specific module. Previous studies have used the indicators developed for adults and applied them to pediatric patients. Furthermore, because of the data sets used, these studies were unable to categorize use impact in more detail to determine whether these indicators disproportionately affected specific charge categories.
We sought to determine the rates of the AHRQ pediatric-specific quality indicator (PDI) events in tertiary care pediatric hospitals in the United States and to quantify the impact of these events in terms of charges (both overall and according to category) and LOSs.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Data were obtained from the Pediatric Health Information System (PHIS), an administrative database that, during the study period, contained inpatient data from 38 freestanding, academic, not-for-profit, tertiary care pediatric hospitals in the United States. These hospitals are heterogeneous with respect to geographic location, bed size, and average daily census. The hospitals are affiliated with the Child Health Corporation of America (Shawnee Mission, KS), a business alliance of children's hospitals. Participating hospitals provide discharge data including demographics, diagnoses, and procedures. Data are subjected to various reliability and validity checks before being incorporated into the database.
Similar to other administrative databases, PHIS contains all of the data necessary to run the AHRQ PDI software. PHIS data include discharge-level variables such as age, gender, ethnicity, LOS, primary expected payer, disposition, ICD-9-CM diagnoses and procedures (
21 each), and geographic census region. Charges in the PHIS database are adjusted for hospital location using the Centers for Medicare and Medicaid price/wage index; cost data are not available in PHIS. In addition, the PHIS database contains billing data that enable users to examine charges categorized according to pharmacy, laboratory, imaging, clinical, supply, and other. Other charges include room and nursing charges.
The AHRQ algorithms use ICD-9-CM codes in a discharge's primary or secondary diagnosis and procedural fields to determine whether the discharge is "at risk" for a given indicator (ie, in the denominator), as described previously.9 A discharge is considered in the numerator for the given indicator if the discharge is at risk and has a code for the particular event. The algorithms define the numerator and denominator for each indicator very specifically in an effort to ensure that all of the PDI episodes represent true, preventable patient-safety events and that patients at high risk for developing those events as a consequence of their condition are excluded from consideration. For example, a discharge is part of the denominator, or "risk pool," for the PDI "postoperative sepsis" if it was an elective surgical discharge, without a principal diagnosis of infection or any code for immunocompromised state or cancer. The discharge would be in the numerator of this PDI if it were in the denominator and had a code for septicemia in any secondary diagnosis field. Likewise, respiratory diseases are excluded for the postoperative respiratory failure PDI, and LOS <5 days or decubitus ulcer in the primary diagnosis code are excluded for the decubitus ulcer PDI. The denominators for the various postoperative PDIs can, therefore, vary on the basis of the exclusions applied to the overall risk pool. One of the 13 PDIs is an indicator of case volume (pediatric heart surgery volume) and was not included in the analysis, because it did not affect LOS or charges.
Participants
All of the discharges from any of the 38 freestanding children's hospitals participating in PHIS between January 1 and December 31, 2006, were eligible for inclusion.
Analysis
To adjust for confounding variables, methods of matching case subjects (ie, discharges experiencing a particular event) with control subjects (ie, discharged subjects not experiencing that event) have been suggested.10,11 This methodology is also appropriate when there is an extreme imbalance in the number of case and control subjects or when the study design is observational.12 Research has determined that gains in bias reduction can be made by matching each case subject with multiple control subjects.13,14
In this study, we matched each case subject with 3 control subjects within the same all-patient refined diagnosis-related group (APR-DRG [3M Corporation, St Paul, MN]) severity level, age group (as defined by the American Academy of Pediatrics as <30 days, 30–364 days, 1–4 years, 5–12 years, 13–17 years, and 
18 years), and hospital. Per the AHRQ software, the APR-DRG risk of mortality level and the risk-adjusted classification for congenital heart surgery score, instead of the APR-DRG severity level, were used for 1 indicator (in-hospital mortality after heart surgery).15 If >3 control subjects were available on the basis of these restrictions, we used propensity scores to minimize the bias in selecting matched control subjects.16,17 Propensity score analysis attempts to identify patients who are similar except for their case or control status; propensity scores reflect the probability of being a case subject on the basis of the patient's observed covariates.18 The propensity score analysis was conducted as follows. The probability (ie, the propensity score) that any patient would be a case subject (ie, have a potentially harmful event) during hospitalization was estimated using a multivariable logistic regression model that incorporated the following patient characteristics: primary payer, gender, disposition, and race. Patients with missing values for any component of the propensity score calculation were excluded. Nearest-neighbor matching was used to select 3 control subjects with the closest propensity scores to the case subject's propensity score.16 Once a case subject was matched with 3 control subjects, the control subjects were removed from the total risk pool.
The difference in use was then computed as the difference between each individual case subject's use and the mean use of the (3) matched control subjects. The mean and SE of these differences were reported. Statistical significance for the difference in use between the case and control subjects was determined by using Wilcoxon's signed rank test, a nonparametric alternative to the 1-sample t test.
Case subjects who could not be matched to a control subject were removed from the analysis. To assess whether bias occurred in the matching process and how such bias would affect our interpretation of results, we compared matched and unmatched case subjects for each PDI to evaluate for differences in demographics and outcomes between the 2 groups.
For this study, the AHRQ's SAS 3.1 (March 2007, SAS Institute, Inc, Cary, NC) was used.19 All of the statistical analyses were performed by using the statistical software SAS 9.1 (SAS Institute, Inc). Because 8 comparisons were made on the same sample of discharges (LOS, total charges, and 6 charge categories), we used the conservative Bonferroni correction20 to set the statistical significance at P < .006. This study received an exemption from the institutional review board at the Children's Hospital of Philadelphia.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
There were 431524 discharges in the PHIS database during the study period (Table 1). Most discharges were between 1 and 12 years of age. Non-Hispanic white and non-Hispanic black race were most common, although Hispanic subjects composed 18% of the study population. The largest representation of discharges (40%) was from the southern census region of the United States.
|
Table 2 shows the PDI rates in the PHIS database and the Kids' Inpatient Database (KID). In PHIS, the most frequent PDIs overall were infection because of medical care, postoperative respiratory failure, and postoperative sepsis. However, the highest PHIS rate was for in-hospital mortality after pediatric heart surgery, with 39.57 per 1000 discharges at risk. Although this PDI only occurred 402 times, the at-risk population was the smallest of the reported PDIs. Overall, the rates of PDIs reported by the AHRQ using the KID were similar to the rates reported in PHIS. There were minor differences between the 2 databases in the following PDIs: in-hospital mortality after pediatric heart surgery, postoperative hemorrhage and hematoma, postoperative respiratory failure, and transfusion reactions.
|
Table 3 displays the matching rate, the excess LOS, and total charges based on PDI occurrences noted in the 2006 PHIS data. The matching rate varied across the indicators, with the lowest proportion of matched case subjects for postoperative respiratory failure and the highest for iatrogenic pneumothorax in neonates. Using these 2 PDIs as examples, 398 (37%) of the 1076 total case subjects with postoperative respiratory failure matched
1 control subject; 184 case subjects matched to only 1 control subject, 54 case subjects matched to 2 control subjects, and 160 case subjects matched to 3 control subjects. Likewise, 12 (92%) of the 13 total case subjects with iatrogenic pneumothorax in neonates matched to 1 control subject; 2 case subjects matched to only 1 control subject, 0 case subjects matched to 2 control subjects, and 10 case subjects matched to 3 control subjects.
|
Age was the only demographic factor with any statistically significant differences between matched and unmatched case subjects, with differences between cases of accidental puncture and laceration, postoperative hemorrhage or hematoma, postoperative sepsis, and infection because of medical care. The demographic variables race, gender, payer, disposition, and census region had no differences in any of the PDIs. Differences in LOS, total charges, or any of the charge categories between matched and unmatched case subjects occurred only in the PDIs of postoperative respiratory failure, postoperative sepsis, and infection because of medical care. Unmatched case subjects with postoperative respiratory failure, who were excluded from analysis, had significantly increased total, pharmacy, laboratory, imaging, clinical, other charges, and LOS relative to matched case subjects. The mean total charges for this PDI were $155397 for matched case subjects and $224275 for unmatched case subjects. Unmatched case subjects with infection because of medical care were older than matched case subjects (2.72 vs 0.97 years) and had lower pharmacy ($40908 vs $54256) and laboratory ($29973 vs $37023) charges than matched case subjects. Unmatched case subjects with postoperative sepsis were older than matched case subjects (2.40 vs 0.85 years) and had higher pharmacy charges ($64800 vs $55464).
The occurrence of 6 of the 12 PDIs examined (accidental puncture and laceration, decubitus ulcer, foreign body left in during procedure, postoperative respiratory failure, postoperative sepsis, and infection because of medical care) was associated with a statistically significant increase in LOS. This increase ranged from 2.8 days for accidental puncture and laceration to 23.5 days for postoperative sepsis.
Statistically significant excesses in overall charges were seen for 8 of the 12 PDIs examined (all except foreign body left in during a procedure, iatrogenic pneumothorax for neonates, postoperative wound dehiscence, and transfusion reaction). The significant excess charges ranged from a minimum of $34884 for accidental puncture and laceration to maximums of $261173 for postoperative sepsis and $337226 for in-hospital mortality after pediatric heart surgery.
Table 4 displays the excess charges according to category; 4 PDIs (33.3%) had increases in all 6 of the charge groups. Statistically significant excess pharmacy charges ranged from $7705 (accidental puncture and laceration) to $62012 (postoperative sepsis). Statistically significant supply charges ranged from $1718 (accidental puncture and laceration) to $20905 (in-hospital mortality and pediatric heart surgery). Eight (66.7%) of the PDIs were associated with excess laboratory charges, ranging from $7622 (accidental puncture and laceration) to $78048 (in-hospital mortality and pediatric heart surgery). Excesses in imaging charges were the most moderate of the 6 categories, ranging from $1956 (accidental puncture and laceration) to $11794 (in-hospital mortality and pediatric heart surgery). Excesses in clinical charges ranged from $10929 for postoperative respiratory failure to $35128 for postoperative sepsis. Eight (66.7%) of the PDIs were associated with excess other charges (ie, room and nursing), from a minimum of $11094 for accidental puncture and laceration to a maximum of $97805 for postoperative sepsis.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
This is the first study to use PDIs to assess the association between patient-safety events in children and resource use. Our exploration of the relationship between PDIs and individual charge categories, total charges, and LOS refines our understanding of these associations. In an era in which quality measures and benchmarking are becoming increasingly important, these indicators provide a rational place to focus quality improvement efforts seeking to reap significant patient and cost benefits. On the basis of our data, efforts to eliminate postoperative sepsis and infection because of medical care could create a potential savings of $722119958 and 83901 patient-days of in-hospital care annually. Creating national standards and benchmarks for these indicators would allow individual pediatric institutions to target the specific patient-safety measures that would bring the greatest benefit to their own patients and payers and would permit institutions with the lowest PDI rates to share their best practices with other institutions nationwide. The drive to reduce these increased charges is even more timely given the recent decision by the Centers for Medicare and Medicaid Services to deny hospitals reimbursement for certain adverse events, such as foreign body left in during procedure, decubitus ulcers, and certain infections because of medical care.21,22
Our study differs from previous studies by using PDIs rather than the adult-derived PSIs. Some PSIs have unclear relevance to children and have been exceedingly uncommon in the pediatric population.7 For example, death in low-mortality diagnosis-related groups and in-hospital postoperative hip fracture both occurred in <0.5 per 1000 pediatric discharges.7 In contrast, events in the newly added PDIs were common in PHIS, ranging from 1.2 per 1000 discharges for iatrogenic pneumothorax in a neonate to >39 per 1000 discharges for in-hospital mortality after pediatric heart surgery. These data suggest that the PDIs function as intended by serving as more appropriate indicators of adverse patient-care events occurring in children. Furthermore, many PSIs not included as PDIs have had minimal contributions to pediatric excess health resource use. Miller and Zhan7 reported no significant increase in LOS or total charges for anesthesia complications, in-hospital postoperative hip fracture, and obstetric trauma. The contributions of the newly added indicators to resource use in our study were variable but at times substantial: in-hospital mortality after pediatric heart surgery contributed to more than $300000 in excess total charges per event.
Many of the 12 PDIs examined reflect overall in-hospital care, but 3 directly reflect quality of care at the provider level. The 2 iatrogenic pneumothorax indicators, which reflect the quality of physician care, were not associated with an increased LOS, and only iatrogenic pneumothorax in nonneonates was associated with increased total, laboratory, supply, imaging, and other charges. Neonates who experienced an iatrogenic pneumothorax may have already been so sick that experiencing this event did not significantly affect their hospitalization, but older children may have experienced an iatrogenic pneumothorax during a routine procedure, such as placement of a subclavian or internal jugular venous catheter, and this complication may have added a significant new dimension to their hospitalization. The third indicator reflecting provider-level care is decubitus ulcer, which can be seen as "nursing-sensitive"23 in that rates of decubitus ulcers are influenced primarily by the quality of care provided by the bedside nurse. We found decubitus ulcers to be associated with a significant increase in LOS and increased overall charges, primarily clinical and other charges. Monitoring decubitus ulcer rates might allow institutions to determine best nursing practices and decrease the overall LOS for the patient population most at risk for this complication.
The remaining PDIs reflect adverse patient-safety events stemming not from provider-level errors but from errors in systems of care. An area in which systems-based issues are of particular importance is the postoperative setting. As a group, LOS and charges for the postoperative PDIs were the most increased of the PDIs. In particular, postoperative respiratory failure and postoperative sepsis were associated with increases in all 6 of the charge categories, and postoperative hemorrhage or hematoma was associated with increases in 4. Together these postoperative complications incur approximately $350 million in charges in PHIS-participating hospitals annually. Concerted effort to reduce the occurrence of these postoperative complications has the potential to remove significant stresses from the health care system. In particular, in-hospital mortality after pediatric heart surgery was associated with increased charges but not increased LOS. By its very nature, in-hospital mortality affects overall LOS but could occur at any point in a patient's hospital course and, thus, should not affect LOS consistently when compared with patients surviving to discharge. The patients who ultimately expired may have been composed of a sicker group than their surviving counterparts, although we matched patients to control subjects with similar APR-DRG severity levels, ages, and hospitals to control for this possibility. Similarly, care teams may have gone to great lengths to prevent mortality, thereby consuming many more resources during these patients' hospitalizations. Pediatric heart surgery patients are likely not unique in this regard, however; Odetola et al24 showed a 2.5-fold increase in hospital charges for nonsurvivors of severe pediatric sepsis relative to survivors.
The overall rates of PDIs in PHIS were well matched with those reported by the AHRQ, with 4 minor differences. For 3 indicators (postoperative hemorrhage or hematoma, postoperative respiratory failure, and transfusion reactions), the PHIS rates were minimally higher than those in KID. This finding is not surprising: children with complex health care needs, who tend to receive care in the tertiary care hospitals contained in the PHIS database, may be more prone to medical errors.8,25 For the fourth, in-hospital mortality after heart surgery, the PHIS rate was lower than in KID. This lower rate may also stem from a difference between PHIS and KID. The PHIS data represent freestanding, academic, tertiary care pediatric hospitals, whereas the KID data for this indicator also include pediatric heart surgeries performed in large university hospitals. Historically, pediatric heart surgeries are more commonly performed at children's rather than general hospitals, and centers with a lower annual case volume of pediatric heart surgeries have been shown to have increased mortality rates.26,27 The PHIS rate might be lower than the KID rate for this indicator because the KID data also include surgeries performed in lower volume, large university hospitals.
Our study has 3 significant limitations. First, PDIs use administrative data, based on coding and physician documentation, and not events themselves. Both errors of missed or wrong coding of true events and errors of coding nonevents as events could occur. Because these errors occur inconsistently, we cannot predict how they might affect our results. To date, no studies have published the sensitivity and specificity of the PDIs compared with actual adverse patient-care events. Likewise, given the nature of our study, we are only able to comment on associations between PDIs and charges and LOS and not on causation. The second limitation is that no matching algorithm of case subjects to control subjects creates perfect matching, and differences between matched and unmatched case subjects of each of the PDIs are a potential source of bias, because unmatched case subjects were not included in the analysis. However, only 3 PDIs had significant differences in LOS or any charge categories between matched and unmatched case subjects. Unmatched case subjects with postoperative respiratory failure had significantly higher LOS, total charges, and categorized charges for all of the categories except supply charges, and unmatched case subjects with postoperative sepsis had significantly higher pharmacy charges. Therefore, exclusion of unmatched case subjects would cause us to underestimate the impact of these PDIs on resource use, and their effect on LOS and charges may be even larger than our study suggests. Unmatched case subjects with infection because of medical care had slightly lower pharmacy and laboratory charges, which may have led to an overestimate of charges. However, unmatched case subjects still had substantially increased charges relative to control subjects. Lastly, the PHIS database provides billed charge data rather than cost data, which may overestimate the economic impact of these adverse events, because payers frequently reimburse at rates less than full charges. Charge data can vary according to geographic region, although we adjusted for hospital location using the Centers for Medicare and Medicaid price/wage index to control for this variation.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Our study provides evidence that pediatric-specific quality indicators are associated with increased LOS and overall charges at freestanding, tertiary care children's hospitals. These PDIs more appropriately identify pediatric adverse events than the previously used adult indicators. These 12 PDIs may not all be equally preventable, and tackling some may provide greater health and cost benefits than others. Additional studies able to determine simple methods of even modest charge reductions within individual PDIs have the potential to create significant cost savings for hospitals nationwide.
|
FOOTNOTES |
|---|
Accepted Dec 12, 2007.
Address correspondence to Samir S. Shah, MD, MSCE, Children's Hospital of Philadelphia, Division of Infectious Diseases, Room 1526 (North Campus), 34th Street and Civic Center Boulevard, Philadelphia, PA 19104. E-mail: shahs{at}email.chop.edu
The authors have indicated they have no financial relationships relevant to this article to disclose.
| What's Known on This Subject PSIs serve as a tool using administrative data to identify potentially preventable adverse patient-care events such as an iatrogenic pneumothorax or a foreign body left in during a procedure. The PSIs were initially tested and validated in adults. Recognizing that PSIs derived for adults were not always applicable to children, the AHRQ revised the PSIs developed for the adult population to create a pediatric-specific module.
|
| What This Study Adds This is the first study to use pediatric-specific PSIs to assess the association between patient-safety events in children and resource use. In an era in which quality measures and benchmarking are becoming increasingly important, these indicators provide a rational place to focus quality improvement efforts seeking to reap significant patient and cost benefits.
|
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
- Romano PS, Geppert JJ, Davies S, Miller MR, Elixhauser A, McDonald KM. A national profile of patient safety in U.S. hospitals.
Health Aff (Millwood). 2003;22
(2):154
–66
[Abstract/Free Full Text] - Miller MR, Elixhauser A, Zhan C, Meyer GS. Patient safety indicators: using administrative data to identify potential patient safety concerns. Health Serv Res. 2001;36 (6 pt 2):110 –132[Medline]
- Rosen AK, Rivard P, Zhao S, et al. Evaluating the patient safety indicators: how well do they perform on Veterans Health Administration data? Med Care. 2005;43 (9):873 –884[CrossRef][Web of Science][Medline]
- Rosen AK, Zhao S, Rivard P, et al. Tracking rates of patient safety indicators over time: lessons from the Veterans Administration. Med Care. 2006;44 (9):850 –861[CrossRef][Web of Science][Medline]
- Coffey RM, Andrews RM, Moy E. Racial, ethnic, and socioeconomic disparities in estimates of AHRQ patient safety indicators. Med Care. 2005;43 (3 suppl):I48 –I57[Medline]
- Zhan C, Miller MR. Excess length of stay, charges, and mortality attributable to medical injuries during hospitalization.
JAMA. 2003;290
(14):1868
–1874
[Abstract/Free Full Text] - Miller MR, Zhan C. Pediatric patient safety in hospitals: a national picture in 2000.
Pediatrics. 2004;113
(6):1741
–1746
[Abstract/Free Full Text] - Miller MR, Elixhauser A, Zhan C. Patient safety events during pediatric hospitalizations.
Pediatrics. 2003;111
(6 pt 1):1358
–1366
[Abstract/Free Full Text] - Sedman A, Harris JM 2nd, Schulz K, et al. Relevance of the Agency for Healthcare Research and Quality Patient Safety Indicators for children's hospitals.
Pediatrics. 2005;115
(1):135
–145
[Abstract/Free Full Text] - Rubin D. Matching to remove bias in observational studies. Biometrics. 1973;29 (1):159 –183[CrossRef][Web of Science]
- Rosenbaum PR. Optimal matching in observational studies. J Am Stat Assoc. 1989;84 (408):1024 –1032[CrossRef][Web of Science]
- Rubin D. Using multivariate matched sampling and regression adjustment to control bias in observational studies. J Am Stat Assoc. 1979;74 (366):318 –328[CrossRef][Web of Science]
- Smith HL. Matching with multiple controls to estimate treatment effects in observational studies. Sociol Methodol. 1997;27 (1):325 –353[CrossRef][Web of Science]
- Ming K, Rosenbaum PR. Substantial gains in bias reduction from matching with a variable number of controls. Biometrics. 2000;56 (1):118 –124[CrossRef][Web of Science][Medline]
- Jenkins KJ. Risk adjustment for congenital heart surgery: the RACHS-1 method. Semin Thorac Cardiovasc Surg Pediatr Card Surg Annu. 2004;7 :180 –184[Medline]
- Rosenbaum PR, Rubin DB. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat. 1985;39 (1):33 –38[CrossRef]
- Joffe MM, Rosenbaum PR. Invited commentary: propensity scores.
Am J Epidemiol. 1999;150
(4):327
–333
[Abstract/Free Full Text] - D'Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 1998;17 (19):2265 –2281[CrossRef][Web of Science][Medline]
- AHRQuality Indicators. Pediatric quality indicators download. Available at: www.qualityindicators.ahrq.gov/pdi_download.htm. Accessed October 18, 2007
- Miller RG Jr. Simultaneous Statistical Inference, 2nd ed. New York: Springer Verlag; 1981
- Lubell J. CMS: your mistake, your problem—eight hospital-acquired conditions won't be paid for. Mod Healthc. 2007;37 (33):10 –11[Medline]
- Pear R. Medicare says it won't cover hospital errors. New York Times. August 19, 2007; Front page
- National Quality Forum. National Voluntary Consensus Standards for Nursing-Sensitive Care: An Initial Performance Measure Set. National Quality Forum Consensus Report. Washington DC: National Quality Forum; 2004
- Odetola FO, Gebremariam A, Freed GL. Patient and hospital correlates of clinical outcomes and resource utilization in severe pediatric sepsis.
Pediatrics. 2007;119
(3):487
–494
[Abstract/Free Full Text] - Slonim AD, LaFleur BJ, Ahmed W, Joseph JG. Hospital-reported medical errors in children.
Pediatrics. 2003;111
(3):617
–621
[Abstract/Free Full Text] - Jenkins KJ, Newburger JW, Lock JE, Davis RB, Coffman GA, Iezzoni LI. In-hospital mortality for surgical repair of congenital heart defects: preliminary observations of variation by hospital caseload.
Pediatrics. 1995;95
(3):323
–330
[Abstract/Free Full Text] - Hannan EL, Racz M, Kavey RE, Quaegebeur JM, Williams R. Pediatric cardiac surgery: the effect of hospital and surgeon volume on in-hospital mortality.
Pediatrics. 1998;101
(6):963
–969
[Abstract/Free Full Text]
PEDIATRICS (ISSN 1098-4275). ©2008 by the American Academy of Pediatrics
CiteULike 
Connotea 
Del.icio.us 
Digg 
Facebook 
Reddit 
Technorati 
Twitter What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
