Objective. Our objective was to describe potential patient safety events for hospitalized children, using the patient safety indicators (PSIs), and examine associations with these events.
Methods. PSI algorithms, developed by researchers at the Agency for Healthcare Research and Quality to identify potential in-hospital patient safety problems using administrative data, were applied to 3.8 million discharge records for children under 19 years from 22 states in the 1997 Healthcare Cost and Utilization Project. Prevalence of PSI events and associations with patient-level and hospital-level characteristics, length of stay, in-hospital mortality, and total charges were examined.
Results. The prevalence of pediatric patient safety events is significant with the highest rate found for birth trauma at 1.5 cases per every 100 births. The majority of these events for birth trauma consist of long bone and skull fractures, excluding the clavicle. Compared with records without PSI events, discharges with PSI events had 2- to 6-fold longer lengths of stay, 2- to 18-fold higher rates of in-hospital mortality, and 2- to 20-fold higher total charges. Bivariate and multivariate analyses found that all PSI events except birth trauma were directly associated with factors related to greater severity of illness and large urban teaching institutions. Birth trauma, however, was directly associated with black and Hispanic ethnicity but was not consistently associated with technologically sophisticated teaching institutions.
Conclusions. The prevalence of birth trauma and other potential patient safety events for hospitalized children is high and comparable to hospitalized adults. These events are associated with increased length of stay, in-hospital mortality, and total charges. Associated factors differ significantly for birth trauma compared with other PSI events. Institutional application of the PSIs may be useful to identify processes of care that warrant further evaluation as the health care industry tackles the problem of patient safety, particularly for children.
The Institute of Medicine (IOM) Report To Err Is Human, released in 1999, shined a spotlight on preventable medical errors.1 The second IOM report, Crossing the Quality Chasm: A New Health System for the 21st Century, reinforced that patient safety is an important goal of our health care system.2 At the National Summit on Medical Errors and Patient Safety Research, held September 11, 2000, Dr J. Eisenberg, former Director of the Agency for Healthcare Research and Quality (AHRQ), likened the problem of medical errors to an epidemic, and noted that we are currently in the first stages of understanding this epidemic. Logistically, this means that research is necessary to understand the magnitude of the problem, its causes, and its burden on patients in the United States.
In response to the first IOM report, AHRQ’s reauthorization by Congress in 1999 clearly stated that a goal of AHRQ was to reduce errors in medicine by 1) identifying the causes of preventable health care errors and patient injury in health care delivery; 2) developing, demonstrating, and evaluating strategies for reducing errors and improving patient safety; and 3) disseminating effective strategies throughout the health care industry. With this mandate in mind and an understanding that patient safety initiatives are designed to prevent adverse outcomes from medical errors, a team of intramural researchers at AHRQ developed a set of patient safety indicators (PSIs) for identifying potential instances of clinically significant compromised patient safety in the inpatient setting.3,4 The PSIs were specifically and conservatively created to target events with a high likelihood of representing errors, such as foreign bodies left during procedures. Similar to AHRQ’s Healthcare Cost and Utilization Project Quality Indicators, the PSIs rely on administrative data and, as such, are indicators, not definitive measures, of patient safety concerns.5 The intent of these indicators is to provide a useful screening tool to identify processes of care that warrant further institutional-level evaluation from the patient safety perspective. The PSIs would enable institutions to identify a manageable number of medical records for closer scrutiny.
To date, the PSIs have undergone testing using the 1997 New York State Inpatient Database.3 This study showed that PSI events have a substantial prevalence and that PSI events are associated with excess length of stay, in-hospital mortality, and charges. What is missing in this evaluation, however, is an understanding of the applicability of the PSIs and prevalence of events for pediatric patients.
Although a fair amount of research to date has been done in the area of medical errors and children, the focus has been on medication errors and regional or institutional errors in the emergency department or intensive care unit setting.6–16 None of these studies provide broad-based estimates of the magnitude, scope, and clinical severity of the problem of patient safety for hospitalized children, who are not only at high risk for patient safety events but also at high risk for clinically significant events. One of the criticisms of a recent study on medication errors and children was that there was no description of “the clinical relevance of any of these errors, or whether any… actually resulted in morbidity.”17
Children are subject to unique vulnerabilities that may predispose them to higher rates of in-hospital patient safety events than the adult population. For example, children have a near universal hospitalization for birth, are not able to directly question their own care, and may not have parents or guardians continuously at the bedside to oversee their care. The aim of this project is to provide a near national picture of the epidemiology of PSI events for children using data from 22 states and to explore the relationship between PSI events and in-hospital length of stay, mortality, and charges, and examine the correlates of PSI events.
A complete description of the development and initial testing of the PSI algorithms has been published.3 The development of the PSIs were based on the definition of patient safety in the To Err Is Human report which is freedom from accidental injury because of medical care, or medical errors. Medical errors were further defined, using the definition from the federal response to the IOM report entitled Doing What Counts, as “the failure of a planned action to be completed as intended or use of a wrong plan to achieve an aim. Errors can include problems in practice, products, procedures, and systems.”1,18 This definition excludes acts that did not achieve their desired outcomes, as long as they were not the result of negligence, outcomes caused by the intrinsic properties of the underlying illness or additional patient comorbidities, and outcomes known to be risks of specific procedures. These exclusions distinguish the PSIs from other indicator systems designed to detect complications of care. For example, although sepsis is clearly a complication and, at times, an adverse event caused by medical care, it is not possible to unambiguously define sepsis as a medical error using administrative data because of the unclear underlying issues such as immunocompromised status, timing of onset, and even the definition of sepsis itself.19
Table 1 defines the PSI algorithms using the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes.20 It is important to note the overall conservative nature of the PSIs in trying to identify only cases with clear patient safety concerns. For example, the majority of PSIs oriented to detect surgical patient safety issues are limited in scope to only discharge records designated as “elective” surgical cases to exclude cases of urgent and emergent surgery where the distinction between error and patient-related factors is nearly impossible using administrative data.
Analysis of PSI Events
Our analysis used data on 3.8 million hospital discharges for children 18 years of age or younger from 22 states in the 1997 HCUP State Inpatient Databases.21 The included states are 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. First, we examined the rates for each of the PSIs. Second, we examined the relationship between PSI events and length of stay, percent in-hospital mortality, and total charges, compared with patients not experiencing a PSI event. Last, we conducted bivariate and multivariate regression analyses to examine the associations between PSI events and various patient-level and hospital-level characteristics.
Patient-level variables included age, gender, ethnicity, severity of illness as classified by the All-Patient Refined-Diagnosis Related Groups (APR-DRG) system, and primary expected payer.22
Hospital-level variables were identified from the literature and were obtained from either the state databases or by linkage to the American Hospital Association’s Annual Survey of Hospitals Database, Fiscal Year 1997.23–27 These variables were hospital ownership, teaching status (major teaching defined as Council of Teaching Hospitals [COTH] member and/or American Medical Association-approved residency program, nonteaching defined as neither), resident/bed ratio, nursing staff intensity (number of full-time and part-time registered nurses divided by the hospital’s average daily census), nursing expertise (number of full-time and part-time registered nurses divided by number of full-time and part-time registered nurses plus full-time and part-time licensed practicing nurses), hospital location, total number of hospital beds, number of pediatric discharges, and a technologic sophistication index (total number of the following items of equipment or facilities: open-heart surgery, organ transplantation capability, positron emission tomography, level 1 trauma center, and neonatal intensive care unit).
To capture potential biases in detecting PSI events based on coding practices, we also examined the effect of the number of diagnoses codes and procedure codes typically used by an institution.
To examine variations among hospitals attributed to the types of patients cared for, we included 2 variables previously used in the literature to capture patient-level severity of illness: percent of hospital beds in intensive care units and the proportion of inpatient surgical procedures to annual admissions.23,24
Multivariate logistic regression analyses were completed using all of the above patient-level and hospital-level variables with the difference of bundling each hospital’s patient-level APR-DRG scores to reflect the overall distribution of severity at the hospital level.22 Because the PSIs are likely imperfect in identifying only cases of medical errors as opposed to events reflecting patient comorbidities, case mix adjustment was necessary. However, given the “two way street” association between being more severely ill and having a PSI event, person-level severity adjustment is maximally confounded by this association. Therefore, hospital-level severity was included in the regression analyses. To obtain the APR-DRG hospital case mix for patient severity, we first applied the APR-DRG software to the discharge records and determined each individual patient’s severity of illness using the APR-DRG 4-point scale (class 1 = least severe, class 4 = most severe). Because the APR-DRG software relies on ICD-9-CM codes to assign severity scores, we deleted those ICD-9-CM codes from the APR-DRG scoring algorithm that clearly represent in-hospital patient safety events such as “iatrogenic hypotension” and “transfusion reaction.” Next, for each given institution in 1997 HCUP State Inpatient Databases, we determined the distribution of patients among the 4 severity classes for 1997: ie, the percentage of each hospital’s patients in class 1, class 2, class 3, and class 4.
For the multivariate logistic regression analysis, the following variable categories were used as reference values: age 1 to 4 years, female sex, white ethnicity, private insurance, not-for-profit ownership, nonteaching hospital status defined from resident/bed ratio (COTH status/AMA residency variable deleted because of collinearity), low nursing expertise (nurse staff intensity variable deleted because of collinearity), rural hospital location, lower number of hospital beds, lower number of pediatric discharges, lower number of diagnosis or procedures codes recorded by institution, low percentage of beds in intensive care units, low percentage of inpatient surgical volume, and percentage of patients in severity class 1. The technologic sophistication variable was coded as a numerical variable with each institution receiving 1 point for each of the 5 technologies present to a maximal score of 5.
Nonparametric comparisons of medians for length of stay and total charges were done using the Wilcoxon rank-sum tests since these data were not normally distributed. Comparisons of in-hospital mortality and bivariate associations between patient-level and hospital-level characteristics and PSI events were completed using χ2 tests. Multivariate logistic regression analyses yielded odds ratios (OR) and 95% confidence intervals (CI) for experiencing a PSI event compared with not experiencing a PSI event. Analyses were done using SAS (SAS Institute, Inc, Cary, NC) with a significance level of P < .05.28
Overall Rate of PSI Events for Children in 1997
Table 2 presents the rate of PSI events by PSI group for the 3.8 million pediatric discharges in 22 states for 1997. The event rates range from .2 (foreign body left during procedure) to 154.0 (birth trauma) events per 10 000 discharge records. The case numbers for PSI events in PSI groups 1 to 11 substantially exceed the only 829 cases (event rate of 2.2 cases per 10 000 discharge records) detected by using only the medical care oriented ICD-9-CM E-codes, a subset of codes specifically designated to detect “injuries due to external causes.”
Given the significant 1.5 cases per 100 births rate of birth trauma identified, better delineation of the types and magnitude of events is helpful. Specifics of the >36 000 cases include: 65% of cases (N = 23 474) involved skeletal injuries such as fractures of long bones and skull (excluding clavicle); 16% (N = 5654) involved eye damage, liver/testes/vulva hematomas, liver/spleen ruptures, scalpel wounds, and traumatic glaucoma; 10% (N = 3972) involved injury to brachial plexus; and 7.5% (N = 2679) involved subdural and cerebral hemorrhages. Further examination of these rates found that only 4 of the >36 000 cases also listed osteogenesis imperfecta as a diagnosis, only 1 case listed a ICD-9-CM V-code denoting infant prematurity, defined as birth weight <2499 g or gestation <37 weeks, and only 3375 cases of the >36 000 cases listed ICD-9-CM codes describing the infant as “heavy for dates” or a birth weight >4500 g. Approximately 80% of the injuries to large-for-gestational-age infants comprised skeletal fractures other than clavicle and brachial plexus injuries. Approximately 94% of the birth traumas occurred to singleton gestations. Key to the issue of preventability, it is important to note variation in rates. Looking across the 22 states, the overall rate of birth trauma varied by a magnitude of 3 from 49 to 159 cases per 10 000 births. Even more dramatic, the rate of skeletal injuries to long bones and skull in general varied by a magnitude of 10 from 12 to 111 cases per 10 000 births.
Comparison of Outcomes Based on PSI Events
Analyses of the relationship between PSI events and median length of stay, percent in-hospital mortality, and median total charges are shown in Table 3. Across all 12 PSI groups, discharges with PSI events had 2- to 6-fold greater median length of stay (all 12 P values <.0001). All PSI groups were associated with significantly greater in-hospital mortality rates except suture of laceration during elective surgery and foreign body left during procedure. The magnitude of this increased mortality rate ranged from 2- to 18-fold (all 12 P values <.0001). Similarly, all PSI groups except birth trauma had 2- to 20-fold higher median total charges (all 12 P values <.001).
Factors Associated With PSI Events
Bivariate associations between patient-level and hospital-level characteristics and PSI events were examined for all PSI groups individually. PSI groups 1 to 10 were all found to have similar associations and thus were pooled together for multivariate analyses. In summary, PSI events were associated with: increasing age, male gender, white and black ethnicity, increasing patient APR-DRG severity of illness grouping, Medicare insurance, not-for-profit ownership, teaching institutions, high resident-to-bed ratios, high nurse staff intensity, greater nursing expertise, urban institution location, greater bed size, higher technologic sophistication, greater percentage of beds in intensive care units, greater volume of inpatient surgical procedures, and greater number of pediatric discharges. In contrast, the Birth Trauma group had some key differences in the bivariate associations. These differences consisted of birth trauma being associated with black and Hispanic ethnicity, having no insurance or private insurance, both not-for-profit and public institutions, having no residents on staff, having a lower percentage of beds in intensive care units, and having a lower volume of inpatient surgical procedures. Furthermore, birth trauma was unrelated to technologic sophistication of the institution and, although still related to teaching institutions, the magnitude of rate increase between nonteaching and teaching institutions dropped from a 3.5-fold increase for PSI groups 1 to 10 to a 1.2-fold increase for birth trauma. Because of these differences, the subsequent regression analyses were conducted separately for the Birth Trauma group. Because of the less consistent coding of E-codes, the relatively rare number of events identified by E-codes, and the fact that E-codes are meant to be used in conjunction with the other ICD-9-CM codes on discharge records, we excluded the E-codes from our pooled data on PSI groups 1 to 10.
Multivariate logistic regression analysis, including APR-DRG-derived case-mix adjustment for each institution, was performed using all the variables from the bivariate analyses. The results are shown in Table 4. It should be noted that similar regression equations completed using patient-level severity of illness had findings completely consistent with our chosen analysis using hospital-level adjustment. Initial correlational analyses found collinearity between the 2 unique teaching status variables and the 2 nurse staffing variables. As such the COTH/residency teaching variable and nurse staff intensity variable were omitted from the regression analyses since both were felt to be less salient a measure than their counterparts, namely resident-to-bed ratios and nursing expertise. Overall, these models had moderate predictive value with a “c” statistic of .78 and .65, respectively, for PSI groups 1 to 10 and birth trauma, reflecting that these administrative data-derived variables alone performed moderately at predicting PSI events. For PSI groups 1 to 10 the variables with the greatest predictive value for experiencing a PSI event were Medicare primary insurance, being almost any age other than 1 to 4 years old except for newborns, and those variables that tend to describe large urban teaching institutions, namely having a greater volume of inpatient surgery, being in a major teaching hospital, and having more hospital beds in intensive care units. In contrast, for birth trauma the variables with the greatest predictive value were black ethnicity and having greater nursing expertise. The majority of variables that tend to define large urban teaching institutions were inversely associated with birth trauma (eg, major teaching institutions, greater percentage of beds in intensive care units, greater volume of inpatient surgical procedures). Comparing both regressions solely on patient characteristics, it is noteworthy that while black ethnicity is inversely associated to events in PSI groups 1 to 10, black ethnicity is directly associated with birth trauma.
Our analyses show that patient safety events for hospitalized children occur in high numbers, comparable to hospitalized adults, and that these events have significant associations with increased length of stay, in-hospital mortality, and total charges, again comparable to adults.3 Looking only at birth, the rate of birth trauma exceeds 1 case in every 100 births. Although several other studies have reported on specific types of birth trauma, this study is the first one to take a comprehensive look at many types of birth trauma in a large sample of infants.29–36 Furthermore, factors associated with PSI events differ significantly for birth trauma compared with other PSI events. In general, research on care around the process of birth is difficult because of lack of data linkages between maternal and child records, hospital record name changes from birth to subsequent weeks, and the clear hand-off of care that occurs from the obstetrician and hospital to the community providers. Beyond birth trauma, our results highlight unacceptably high occurrence rates of significant potential patient safety events for children. The majority of events detected by these PSIs occurred to several hundred children per type of event, such as 639 children experiencing a transfusion reaction and 469 children experiencing an obstetrical patient safety event. As the nation’s focus on patient safety continues to increase, we need to assure that reporting systems, educational programs, and team training activities actively involve providers of children’s health care given these results. In addition, our results highlight the need for more focused work, particularly patient safety related, during the critical time period of birth as well as for all hospitalized children. It is important to note that the PSI algorithms are currently being refined and expanded with extensive national expert panel input and will retain the birth trauma grouping as a major patient safety event.
An equally important finding of this work speaks to the utility of the PSIs. Given the main intent of the PSIs to serve as an internal institutional screening tool to identify records for scrutiny with respect to patient safety, these results show that the PSIs may be a useful tool to identify a manageable number of records with high likelihood of containing information pertinent to institutional quality improvement efforts.
In light of the first IOM report, these findings regarding pediatric hospitalizations are not surprising.1 The growing pool of literature that typically focuses on adult patients supports these findings as well. For example, 1 study looking at adult intensive care units reported that, on average, there were 178 activities per patient per day with 1.7 errors per patient per day.37 Also of little surprise is our substantiation of the belief that reliance on only ICD-9-CM E-codes to identify patient safety events from administrative data are insufficient and likely substantially underestimates the problem of patient safety events for both children, as show here, and adults.3
Although it would be desirable to identify causal factors related to the occurrence of PSI events, our analyses do not support direct causal links between any of the patient-level or hospital-level factors and PSI events. Instead, our results highlight patient and system commonalities that predispose one to PSI events caused by underlying similar processes and risk factors. For example, teaching institutions do not purposefully cause PSI events but systems factors common to teaching institutions, such as multiple caregivers, high patient acuity, and use of newer technologies, likely do predispose one to PSI events. What is clear from this work is that the commonalities differ for birth trauma compared with other PSI events such as transfusion reactions and infections attributed to procedures. Given the clear and likely bidirectional association between severity of illness and PSI events, this finding is consistent with the fact that while many institutions may provide birthing services, many fewer provide services to children with chronic higher severity illnesses. Furthermore, these institutions are systematically different and tend toward larger urban teaching hospitals. Better clinical data for case mix adjustment is needed to further explore events relative to these institutions. Better databases able to link up maternal and infant records are also needed to further explore factors specifically associated with birth trauma. Although several other studies have focused on mode of delivery as a risk factor for birth trauma, current administrative data does not permit insight into type of delivery for infants with birth trauma.33–35,38,39
The PSIs rely on hospital administrative data. This is their strength and their most significant limitation. Electronic administrative data are a readily available and relatively inexpensive source of information on a care setting that is particularly susceptible to patient safety concerns because of the intensity of the interventions that occur there. However, administrative data provide limited clinical information and have known problems in coding accuracy, coding variation, limited ability to risk adjust, and limited insight into timing of events.40–43 As 1 example of this, within the state with the highest and clearly outlier rate of birth trauma, analysis of institution-specific rates found several outliers with 1 in particular recording a 66% rate of skeletal injuries among newborns in 1997. A second limitation of the PSIs is the infrequency of the events that are, almost by definition, relatively rare. This issue underscores the appropriate use of the PSIs as institutional case-finding tools aimed at internal quality improvement as opposed to use for directly comparing individual institutions especially in public reports that identify individual hospitals. Third, despite best efforts, the PSIs are likely imperfect in truly identifying only cases of compromised patient safety. This imperfection encompasses both false-positives and false-negatives. As long as the PSIs are used as an internal case finding tool, however, this limitation is acceptable. Last, the PSIs are not a comprehensive catalog of all medical errors that can occur in the hospital setting. Instead the PSIs are intended to be a conservative list of potential errors amenable to detection with administrative data. Because few if any such tools currently exist, particularly for children, this conservative short list can serve as a starting point for understanding and tackling the patient safety problem while better reporting systems are in development. Indeed one may hope that any eventual ideal reporting system may involve triangulation between administrative data, chart review, and voluntary self reports of critical incidents to maximize the ability to identify events. The primary differences between these types of data collection are worth noting as one thinks about creating an ideal system for reporting patient safety events. Administrative data, as analyzed here, are inexpensive, nearly universal, and permit unsolicited identification of potential events although the depth of clinical information is limited. Chart review, on the other hand, provides in-depth clinical information but is fairly expensive to implement on a large scale. Lastly, voluntary critical incident reporting depends completely on the compliance of providers with reporting but does provide real-time in-depth clinical insights. Interdigitated use of these types of data collection would create a system with significant scalability and flexibility in terms of resource needs and provider compliance.
Despite these limitations of the PSIs, the potential benefits need to be delineated. First, the PSIs can have value both cross-sectionally and over time within a defined system of care for institutional quality improvement efforts. Second, PSIs can have value when applied broadly at the state or regional level to provide an initial assessment of the potential scope of the “epidemic” of patient safety events as brought to fruition in our analyses with the identification of >36 000 cases of birth trauma. Third, the PSIs provide a tool to identify cases of errors that result in either morbidity or mortality as opposed to the more traditional tools using only mortality data. Although the debate regarding whether patient safety efforts should focus on injuries or errors is still open, the fact that children may not be as predisposed to death after having a patient safety event as compared with adults calls for a comprehensive approach involving both injuries and errors.44,45 The downside of not using a comprehensive approach for children would be a true case of throwing the baby out with the bath water. Focusing on birth trauma and thinking of patient safety events as risk factors for morbidity, 1 recent study reported that 34% of infants who suffer a brachial plexus injury do not experience full recovery and 14% have severe permanent weakness.46 Last, and perhaps most importantly, the PSIs for children can begin to identify new areas in great need of future work related to patient safety. For example, the large number of birth trauma cases and the broad variability in rates comparing state to state raises questions about the preventability of these significant events and points to an area in great need of further study.
Looking toward the future, AHRQ is currently working under contract with the University of California-Stanford Evidence-Based Practice Center to evaluate and expand this pool of PSIs, obtain clinician input on indicator validity and usefulness, and conduct empirical analyses of the variability of event rates. It is anticipated that there will be a public release of the PSIs, as a component of the AHRQ Quality Indicators, in late 2002.47 Following this development effort, the measures will be explored further and validated to assess the extent to which these are preventable events. With the national attention on patient safety and the improvements in medical information coding, it is anticipated that the PSIs will be an evolutionary list of patient safety events.
Given the potential benefit of an easy tool to identify records likely to include medical error and the magnitude of the issue for children particularly with respect to birth trauma, we wanted to report our findings before the final development of the PSIs and extensive validation work. Almost regardless of the eventual results of any validation work using chart data it is unacceptable to not want to delve deeper into this area to understand and institute prevention especially for newborns. Given the conservative nature of the PSIs to identify cases with high likelihood of error, current use of the PSIs before formal studies comparing them to chart data can result in 1 or both of 2 beneficial outcomes. First, it is likely that true errors will be identified from the records flagged by the PSIs and, hopefully, lead to quality improvement efforts. Second, coding practices at institutions will come under greater scrutiny so that the coding of medical information will become more accurate.
The PSIs are a set of intuitive, administrative data-based indicators of potential patient safety events. They are appropriate for internal quality improvement efforts but are not intended for purchasing decisions, sanctioning individual institutions, or public reporting. Preliminary analyses using the PSIs on data from hospitalized children show that children experience significant numbers of patient safety events, that these event rates are comparable to hospitalized adults, and that attention clearly needs to be paid to the unique event of childbirth. With time, improved measures with better precision, validity, and reliability will likely be developed. In the interim, these indicators and these analyses open the door to many areas apparently ripe for investigation from a patient safety perspective. Perhaps more than ever we are becoming aware that “there are some patients whom we cannot help; there are none whom we cannot harm.”48
This project was funded by intramural research funds of the AHRQ. There are no financial disclosures to make relative to this work.
We would like to acknowledge the support of Ann R. Marshall, MSPH, during her internship at AHRQ with respect to the background literature synthesis. Dr Miller initiated and completed the analysis described here while serving as the Acting Director of the Center for Quality Improvement and Patient Safety at the AHRQ.
- Received June 12, 2002.
- Accepted November 22, 2002.
- Address correspondence to Marlene R. Miller, MD, MSc, FAAP, Director of Quality and Safety Initiatives, Johns Hopkins Children’s Center, CMSC 2-125, 600 N Wolfe St, Baltimore, MD 21287. E-mail:
The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article should be construed as an official position of the AHRQ or the US Department of Health and Human Services.
- ↵Institute of Medicine. To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press; 1999
- ↵Committee on Quality of Health Care in America, Institute of Medicine, ed. Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC: National Academy Press; 2001
- ↵Doing what counts for patient safety: federal actions to reduce medical errors and their impact. Report of the Quality Interagency Coordination Task Force (QuIC) to the President, February 2000
- Institute for Safe Medication Practices and the Pediatric Pharmacy Advocacy Group. Hospital survey shows much more needs to be done to protect pediatric patients from medication errors. In: ISMP Medication Safety Alert; 2000. Available at: www.ismp.org. Accessed April 3, 2003
- Folli HL, Poole RL, Benitz WE, Ruso JC. Medication error prevention by clinical pharmacists in two children’s hospitals. Pediatrics.1987;79 :718– 722
- ↵Quality Interagency Coordination Task Force. Doing What Counts for Patient Safety: Federal Actions to Reduce Medical Errors and Their Impact. Washington, DC: Quality Interagency Coordination Task Force; 2000
- ↵Pronovost PJ, Angus DC, Miller MR. Design considerations in clinical trials on acutely ill patients. Res Practitioner.2000;1 :47– 54
- ↵Public Health Service and Health Care Financing Administration. International Classification of Diseases, 9th Revision, Clinical Modification, Volumes 1, 2, and 3. 8th ed. Washington, DC: Public Health Service; 1997
- ↵http://www.ahrq.gov/data/hcup. Accessed April 3, 2003
- ↵All-Patient Refined Diagnosis Related Groups (APR-DRGs), Version 15.0. Using 3M APR-DRGs: A Guide for Implementing Severity Adjusted Data. Murray, UT: 3M Health Information Systems; 1999
- ↵SAS Institute, Inc. SAS/SAT User’s Guide, Version 6. 4th ed. Cary, NC: SAS Institute, Inc; 1989
- ↵Schwartz RM, Gagnon DE, Muri JH, Zhao QR, Kellogg R. Administrative data for quality improvement. Pediatrics.1999;103 :291– 301
- ↵http://www.ahrq.gov/data/hcup/qinext.htm. Accessed April 3, 2003
- ↵Strauss MB. Familiar Medical Quotations. Boston, MA: Little, Brown & Co; 1968:637a
- Copyright © 2003 by the American Academy of Pediatrics