Objective. To determine the incidence of preventable adverse events (AEs) and near misses (NMs) among infants hospitalized for bronchiolitis at a pediatric tertiary care hospital and the impact of these errors on hospital length of stay (LOS).
Methods. We studied 143 infants with bronchiolitis, ages 0 to 12 months, admitted from December 2002 to April 2003. Using prospective chart review and staff reports, we captured medical errors and AEs. Each event was classified as a (1) preventable AE, (2) nonpreventable AE, (3) intercepted NM, (4) nonintercepted NM, or (5) error with little or no potential for harm.
Results. Of 143 patients, 15 (10%) suffered an AE or NM. The incidence of preventable AEs was 10 per 100 admissions. We found a higher incidence of preventable AEs and NMs among critically ill patients (CIPs) compared with non-CIPs (68 vs 5 per 100 admissions, respectively), making the absolute risk of an AE or NM 14 times more likely in CIPs. Mean LOS was significantly longer for CIPs with at least 1 AE (9.1 ± 8.8 days) than for CIPs without AEs (2.9 ± 1.5 days). Mean LOS was not significantly different between non-CIPs who did (3.8 ± 2.6 days) and did not (4.2 ± 5.0 days) experience an AE.
Conclusions. Preventable AEs occur frequently among patients admitted for bronchiolitis, especially those who are critically ill. CIPs who suffer AEs during their hospitalization have longer hospital LOSs. Future studies should investigate error-prevention strategies with a focus on those patients with severe disease.
Pediatric inpatients frequently suffer medical errors, which may cause severe harm. Although the numbers are debated,1–3 the 1999 Institute of Medicine report estimated that between 44000 and 98000 hospital-related deaths occur each year in the United States as a result of medical errors.4 Although medication error and adverse drug event rates have been reported for pediatric inpatients,5 rates of other medical errors such as diagnostic, preventive, and procedural errors have not been well defined in general or among specific high-risk patient populations.
Bronchiolitis is the leading cause of hospitalization during the first year of life and is a major cause of pediatric morbidity and mortality.6 Patients hospitalized with bronchiolitis, particularly those with severe disease, have been shown to suffer a significant rate of complications.7,8 However, it has remained unclear how many of these complications are inevitable consequences of severe bronchiolitis and how many may be caused by medical errors.
To better understand the nature and consequences of preventable complications among patients hospitalized with bronchiolitis, we conducted a prospective study of infants admitted to a large academic pediatric hospital during a single bronchiolitis season. Our specific aims were to determine the incidence of adverse events (AEs) and near misses (NMs) among these patients and assess the effects of these incidents on hospital length of stay (LOS).
We conducted a prospective study of 143 infants (0–12 months old) admitted for bronchiolitis to a large tertiary care pediatric academic center from December 2002 through April 2003. The inpatient units studied included all general medical wards, a short-stay medical ward, and a 20-bed medical and surgical intensive care unit (ICU). Exclusion criteria were age of >12 months and the presence of another acute respiratory illness such as asthma or bacterial pneumonia on presentation. Critically ill patients (CIPs) were defined as those with severe enough disease on initial presentation to require admission to the ICU. Non-CIPs were all other patients who were initially admitted to the non-ICU inpatient wards. Daily review of patient charts and nursing flow sheets was performed. The institutional review board approved a waiver of consent because of the observational design of this study and the commitment to preservation of patient confidentiality.
General medical ward teams consisted of a pediatric attending physician, a senior pediatric resident, 4 pediatric interns, and 3 to 4 third-year medical students. The short-stay medical team consisted of a pediatric attending physician, a senior pediatric resident, and 2 pediatric interns. Patient-to-nurse ratios ranged from 2:1 to 3:1. The medical/surgical ICU was staffed by an attending pediatric intensivist, 2 pediatric ICU fellows, a senior pediatric resident, and 2 to 3 junior pediatric residents. The patient-to-nurse ratio in the ICU ranged from 1:1 to 2:1. Pediatric clinical pharmacists were present during patient care rounds in the ICU and on the general medical teams. All orders were handwritten by physicians.
The following definitions were used:
Medical error: any error in the delivery of medical care regardless of whether it is harmful or trivial.
Adverse event: any injury resulting from medical management. Injuries resulting from an underlying disease process rather than medical management are not AEs.
Nonpreventable adverse event: unavoidable injury resulting from appropriate, error-free medical care. Nonpreventable AEs are not medical errors.
Preventable adverse event: injury resulting from an error in medical management.
Near miss or potential adverse event: a medical error that has significant potential to cause harm but does not for 1 of 2 reasons:
Intercepted near-miss: a potentially harmful error that is intercepted before reaching the patient.
Nonintercepted near miss: a potentially harmful error that unexpectedly does no detectable harm despite reaching the patient.
Data were collected by a single physician (S.C.M.) from 3 main sources: (1) patient charts and nursing flow sheets; (2) reports submitted by nurses, housestaff, and pharmacists; and (3) hospital administrative and financial databases. A structured standardized form was used to collect information on all patients admitted with bronchiolitis. Age, gender, gestational age at birth, comorbidities, pertinent past medical history, admission date and time, and location of admission (inpatient ward or ICU) were collected. Additionally, clinical data such as a patient's admission weight, chest radiograph findings (if applicable), and medications used within the first 6 hours of admission were extracted from the patient chart and bedside nursing flow sheet. An initial severity score was assigned to each patient based on a validated scoring system for evaluating bronchiolitis patients in the emergency department setting that includes respiratory rate, pulse oximetry, retractions, prematurity, and the presence of atelectasis by chest radiograph.9 Patient admission date, discharge date, and LOS were extracted from the hospital administrative databases.
Reports of possible AEs and medical errors were captured by means of chart and flow-sheet review and reports submitted by nurses, house officers, and pharmacists. With the support of the ICU and medical ward leaders, the cooperation of the medical staff was enlisted before this study's initiation through information sessions with nurses, house officers, and pharmacists that described the nature of incident reporting. Staff reporting forms, which were available in all physician and nursing workrooms on all units, offered a voluntary, anonymous means for staff to submit reports. It was stressed that reporting would not be used to place blame on any individual but to aid in error detection from an epidemiologic and systems-improvement standpoint. The physician reviewer investigated all submitted reports of possible errors. Data were collected 5 days per week. Monday reviews included reviews of charts, reports, and flow sheets from the weekend.
Details of any suspected AE or medical error were recorded by using a structured data-collection form. A description of the incident was recorded along with the time, date, and location of the event. Other data, including the circumstances surrounding the incident and the patient's outcome, were captured to aid in classification.
All suspected incidents were reviewed independently by 2 physicians with experience studying medical errors (D.A.G. and C.P.L.) who classified each as 1 of the following: (1) AE (preventable or nonpreventable); (2) intercepted NM; (3) nonintercepted NM; (4) error with little or no potential for harm; or (5) exclusion (if the incident was felt to represent neither an AE nor an error). The severity of each incident was rated as (1) fatal, (2) life threatening, (3) severe, (4) significant, or (5) not injurious or harmful. Finally, the preventability of an incident was rated as (1) prevented, (2) definitely preventable, (3) probably preventable, (4) probably not preventable, or (5) definitely not preventable. Definitely preventable and probably preventable were collapsed into 1 category (preventable) for analysis, as were probably not preventable and definitely not preventable (not preventable). The type of error was determined to be (1) medication-related, (2) procedural, (3) preventive, or (4) diagnostic. Any disagreements in initial ratings between the independent reviewers were resolved by consensus. κ statistics were calculated for preconsensus ratings as described below.
We report rates of AEs, preventable AEs, and NMs per 100 admissions and per 1000 patient-days. Rates of AEs and NMs per 100 CIP and non-CIP admissions and per 1000 CIP and non-CIP patient-days were compared assuming Poisson distributions. A logistic-regression model was also built to compare CIP and non-CIP incidence rates and to control for differences in demographic variables (age, gender, race, insurance) and prematurity and admission severity-of-illness score.
Differences in LOS between patients with and without AEs were compared by using the Wilcoxon rank-sum test to allow for the nonnormal distributions of LOS. A linear-regression model controlling for age, gender, race, insurance, prematurity, and admission severity-of-illness score was also built.
We calculated preconsensus interrater reliability between physician reviewers by using the percentage of agreement and the κ statistic. Reviewer agreement was excellent for event classification (κ = .71) and preventability (κ = .71). Reviewers had moderate agreement when rating event severity (κ = .40).
This study sample consisted of 143 infants (all <1 year of age) hospitalized for bronchiolitis. The average patient age was 99 days (SD: ±86 days), and 64% were male. The characteristics of the study population as a whole are listed in Table 1. Fourteen percent of the patients were premature. Only 7 of the 143 patients had comorbidities other than prematurity (4 had cardiac disease, 2 had metabolic abnormalities, and 1 had cleft lip).
It is notable that the 19 CIPs and 124 non-CIPs differed significantly in several respects: CIPs were younger (mean age: 60 [CIPs] vs 105 days [non-CIPs]; P = .001), less often male (42% [CIPs] vs 67% [non-CIPs]; P = .04), and more acutely ill (mean severity-of-illness score: 4.8 [CIPs] vs 2.4 [non-CIPs]; P < .0001). In addition, there was a nonsignificant trend toward increased prematurity (26% [CIPs] vs 12% [non-CIPs]; P = .10).
Fifteen (10%) of the 143 patients admitted with bronchiolitis suffered an AE or NM. Four (2.8%) of the 143 patients suffered >1 AE or NM. In total, 26 incidents were detected among the 143 inpatients studied (Table 2). There were 13 preventable AEs, 5 nonpreventable AEs, 2 intercepted NMs, 4 nonintercepted NMs, and 2 errors with little or no potential for harm. The incidence of all AEs was 13 per 100 admissions (29 per 1000 patient-days). The incidence of preventable AEs was 9 per 100 admissions (21 per 1000 patient-days), and the incidence of preventable AEs plus NMs was 13 per 100 admissions (30 per 1000 patient-days). Of the 19 AEs and NMs detected, 1 event was rated as life threatening, 13 were severe, and 5 were significant. Eight were errors in prevention, 5 were diagnostic errors, 4 were procedural errors, and 2 were medication errors. Examples of each type of incident are given in Table 3.
We found a higher incidence of preventable AEs and NMs in CIPs compared with non-CIPs (68 vs 5 per 100 admissions, respectively [P < .001]; 117 vs 12 per 1000 patient-days [P < .001]), making the absolute risk of an AE or NM occurring during hospitalization 14 times more likely among the CIP population. Using multivariate analysis controlling for age, gender, prematurity, race, insurance, and severity-of-illness score, the odds ratio of an AE occurring during a CIP vs a non-CIP admission was 9.5 (95% confidence interval: 2.3–39.3; P = .002).
Unadjusted mean LOS was longer for CIPs who had at least 1 AE (9.1 ± 8.8 days) than for CIPs who did not (2.9 ± 1.5 days) (P = .07). The LOS difference among CIPs who did and did not experience an AE was preserved in multivariate linear-regression analysis, controlling for age, gender, race, insurance, severity of illness, and prematurity (P = .04). Mean LOS among non-CIPs who did (3.8 ± 2.6 days) and did not (4.2 ± 5.0 days) experience an AE was not significantly different (P = .55).
We found that preventable AEs and potentially harmful medical errors occurred in 10% of infants hospitalized with bronchiolitis. Procedural, preventive, and diagnostic errors were most common; medication errors were relatively infrequent. CIPs experienced a higher rate of AEs than non-CIPs, and many experienced >1 AE. Occurrences of AEs in CIPs were associated with a significant prolongation in LOS.
A 2003 study characterizing complications in 684 infants hospitalized for bronchiolitis found that most infants (79%) suffered ≥1 complications.7 Serious complications occurred among 24% of these inpatients. Both serious and minor complications were found to be associated with significantly longer LOS and higher costs. Respiratory complications were most common, including apnea, pneumothorax, hyperinflation, and atelectasis. This retrospective study, however, was not designed to differentiate between complications resulting from therapy and those resulting from the underlying disease process; nor was it designed to determine the preventability of treatment-related events. In contrast, our prospective study was designed specifically to identify AEs and medical errors. Complications resulting from bronchiolitis itself, rather than its treatment, were not included in our study.
Our study differed from most prior pediatric patient-safety work5,10,11 in that we sought to determine the incidence of all medical errors with significant potential to cause harm, both related and unrelated to medications. Ten percent of all patients in our study experienced at least 1 preventable AE or NM. This rate is similar to that reported in some prior studies of AEs in inpatient populations (consisting primarily of adults); the Quality of Australian Health Care Study (QAHCS) found that 16.6% of admissions were associated with an AE.12 Other large studies of AE rates using chart-review methodologies, including the Harvard Medical Practice Study and the Utah Colorado Study (UTCOS), have found rates of AEs to be lower.13,14 The differences in detection rates between these studies and ours are likely a result of differences in study populations and definitions of AEs. Prior work has demonstrated that UTCOS reviewers were less likely to call certain types of problems AEs than were QAHCS reviewers, whose rates and definitions more closely paralleled our own for this study.15,16
Prior pediatric research has been performed to determine the rate of hospital-reported preventable injuries in pediatric inpatients in the United States by using administrative databases and the Agency for Healthcare Research and Quality patient-safety indicators. Rates determined by using these methodologies tend to be lower, because it is only possible to capture those incidents that receive a billing code suggestive of an AE; in addition, it is not possible in such studies to confirm that an AE truly occurred on a case-by-case basis. Slonim et al17 identified 1.81 to 2.96 errors per 100 discharges using International Classification of Diseases, Ninth Revision codes from hospital discharge records. Using the Agency for Healthcare Research and Quality patient-safety indicators tool, Miller and Zhan18 identified suspected AE rates in the range of <1 to 2152 per 10000 discharges across the range of indicators relevant to pediatric populations. Differences in patient populations explain some of the differences in our rates of events and the rates identified by using these methods, but much of the difference almost certainly stems from differences in data-collection methodologies.
Our study, using an intensive prospective surveillance methodology, complements these large pediatric database studies by providing more comprehensive and definitive data on AEs in a focused population. Additional studies of pediatric patient safety using prospective surveillance methodologies will be needed to better elucidate, across a range of discrete pediatric subpopulations, the incidence and consequences of possible AEs identified through large database-mining efforts.
We discovered a greater incidence of preventable AEs in CIPs. Prior studies in adult populations have found CIPs to be more likely to suffer AEs and errors,19 but this finding may be magnified in bronchiolitis for several reasons. The care of bronchiolitis on the medical wards in this study's institution is highly protocolized because of the use of a clinical practice guideline. Because few if any medical therapies have been proven to be of benefit in the treatment of bronchiolitis, ward care consists primarily of noninvasive, supportive therapies. This may explain the smaller proportion of medication errors in this study when compared with errors in diagnosis, procedures, or prevention. In sharp contrast, the small percentage of patients who are sick enough to require ICU-level care often need high-intensity, high-risk therapies such as intubation, mechanical ventilation, and central venous catheter placement. Their underlying disease process may be quite different from those patients with a clinical diagnosis of bronchiolitis who are cared for on the medical wards, leading to the need for high-risk therapies; some may actually have acute respiratory distress syndrome. An increased severity of illness consequently leads to a marked discrepancy in risk exposure between CIPs and non-CIPs that is most likely responsible for the difference in error rates observed. The sickest of these patients, with the most severe comorbidities, may suffer multiple events; 1 patient in this study who was a premature twin suffered 5 events. Our findings suggests that medication errors are not the only area of concern in optimizing error prevention and, at least in some populations, may be less of a concern than other types of errors.
There was a significant increase in LOS for those CIPs who suffered AEs compared with those who did not. Although cost data are not reported here, LOS is the primary driver of hospital costs,20 and these increases in LOS likely increased costs as well.
Our study has several limitations. First, this study was performed in a single academic pediatric hospital during a single bronchiolitis season. These results may not be generalizable to the majority of hospitals in which infants receive care for bronchiolitis. Even with meticulous prospective review of all inpatient charts and nursing flow sheets as well as reports from medical staff, our ability to capture events was limited by the degree to which clinical events were documented by the medical staff. We relied on clinical information at the time of prospective review to guide our detection of potential treatment-related events and were not directly observing patient care as it occurred. Because the medical staff were aware of our study, their performance may have been influenced by the presence of study personnel. This effect, however, would be expected to decrease our rate of capturing events, and thus true event rates may actually be higher.
Additionally, because we sought to broadly capture all medical errors, it was not possible to develop explicit review criteria for all possible events. Instead, consistent with prior studies of medication errors,5 we used an implicit review process to classify events. By definition, implicit review relies to a considerable extent on subjective assessment. Although the reliability of this process was high (κ = .71 for event classification and preventability), it is imperfect. Although many events were clear-cut, reasonable clinicians and investigators might disagree regarding the classification and preventability of others. In addition, consistent with prior studies,5 ratings of event severity were less reliable than ratings of classification and preventability. For example, we judged that an apparent Staphylococcus aureus bacterial pneumonia in an intubated respiratory syncytial virus–positive bronchiolitis patient that appeared on day 3 of hospitalization was related, at least in part, to the placement of an endotracheal tube at admission. It is possible, however, that it might have been a coexisting bacterial infection present on admission that presented later in the hospitalization.
Another limitation of this study was our inability to confidently establish a causative link between LOS and the occurrence of an AE or error in the CIP population. Although the occurrence of an AE in a CIP was associated with longer LOS on multivariate analysis controlling for age, gender, race, insurance, prematurity, and severity of illness, a causal relationship between AEs and LOS cannot be fully established. There is, however, some evidence that increased LOS was a result of the occurrence of AEs. First, in CIPs with AEs, the mean pre-event LOS was only 1.6 days, which suggests that AEs did not occur at higher rates in these patients simply because of longer exposure time. Rather, because long LOSs followed AEs temporally, it is more likely the case that the AEs drove LOS up. In addition, the multivariate predictive model for LOS strongly supports the importance of an AE as a predictor of a prolonged hospital course. The multivariate model that we developed explained a high percentage of variability in LOS among CIPs (r2 = 0.57); when occurrence of an event was removed from the model, the variability explained by the model decreased substantially (r2 = 0.36).
Future efforts aimed at preventing medical errors are important for all pediatric inpatients, but particularly for those patients who are at greatest risk of suffering preventable AEs and errors. Bronchiolitis is the leading cause of hospitalization for infants, and the current study demonstrates that these infants represent a high-risk population.
Preventable AEs occur commonly among inpatients hospitalized with bronchiolitis and are associated with longer hospital LOSs in CIPs. Interventions aimed at reducing preventable AEs are essential and, in the case of bronchiolitis, may be most beneficial in the subpopulation of inpatients with severe disease on admission.
This study was funded by a grant from Medimmune, Inc.
We thank all nursing staff, housestaff, and medical staff for their valuable participation during this study. We also thank Medimmune, Inc for their support of this research.
- Accepted December 21, 2004.
- Reprint requests to (S.C.M.) Department of Medicine, Children's Hospital Boston, 300 Longwood Ave, Hunnewell 131, Boston, MA 02115. E-mail:
No conflict of interest declared.
- ↵Institute of Medicine. To Err Is Human: Building a Safer Health System. Kohn LT, Corrigan J, Donaldson MS, eds. Washington, DC: National Academy Press; 1999
- ↵Leader S, Kohlhase K. Recent trends in severe respiratory syncytial virus (RSV) among US infants, 1997–2000. J Pediatr.2003;143(5 suppl) :S127– S132
- ↵Wilson D, Landrigan C, Horn S, Smout R. Complications in infants hospitalized for bronchiolitis or respiratory syncytial virus pneumonia. J Pediatr.2003;143(5 suppl) :S142– S149
- ↵Prais D, Schonfeld T, Amir J. Admission to the intensive care unit for respiratory syncytial virus bronchiolitis: a national survey before Palivizumab use. Pediatrics.2003;112 :548– 552
- ↵Fortescue EB, Kaushal R, Landrigan CP, et al. Prioritizing strategies for preventing medication errors and adverse drug events in pediatric inpatients. Pediatrics.2003;111 :722– 729
- ↵Folli HL, Poole RL, Benitz WE, Russo JC. Medication error prevention by clinical pharmacists in two children's hospitals. Pediatrics.1987;79 :718– 722
- ↵Thomas EJ, Studdert DM, Runciman WB, et al. A comparison of iatrogenic injury studies in Australia and the USA. I: Context, methods, casemix, population, patient and hospital characteristics. Int J Qual Health Care.2000;12 :371– 378
- ↵Runciman WB, Webb RK, Helps SC, et al. A comparison of iatrogenic injury studies in Australia and the USA. II: Reviewer behaviour and quality of care. Int J Qual Health Care.2000;12 :379– 388
- ↵Slonim A, LaFleur B, Ahmed W, Joseph J. Hospital-reported medical errors in children. Pediatrics.2003;111 :617– 621
- ↵Miller M, Zhan C. Pediatric patient safety in hospitals: a national picture in 2000. Pediatrics.2004;113 :1741– 1746
- Copyright © 2005 by the American Academy of Pediatrics