Abstract
Background
Inpatient medical injuries among children are common and result in a longer stay in hospital and increased hospital charges. However, previous studies have used screening criteria that focus on inpatient occurrences only rather than on injuries that also occur in ambulatory or community settings leading to hospital admission.
Objective
To describe the incidence and outcomes of medical injuries among children hospitalized in Wisconsin using the Wisconsin Medical Injury Prevention Program (WMIPP) screening criteria.
Methods
Cross sectional analysis of discharge records of 318 785 children from 134 hospitals in Wisconsin between 2000 and 2002.
Results
The WMIPP criteria identified 3.4% of discharges as having one or more medical injuries: 1.5% due to medications, 1.3% to procedures, and 0.9% to devices, implants and grafts. After adjusting for the All Patient Refined‐Diagnosis Related Groups disease category, illness severity, mortality risk, and clustering within hospitals, the mean length of stay (LOS) was a half day (12%) longer for patients with medical injuries than for those without injuries. The similarly adjusted mean total hospital charges were $1614 (26%) higher for the group with medical injuries. Excess LOS and charges were greatest for injuries due to genitourinary devices/implants, vascular devices, and infections/inflammation after procedures.
Conclusions
This study reinforces previous national findings up to 2000 using Wisconsin data to the end of 2002. The results suggest that hospitals and pediatricians should focus clinical improvement on medications, procedures, and devices frequently associated with medical injuries and use medical injury surveillance to track medical injury rates in children.
Keywords: medical injuries, hospitalization, children
Safety indicators and chart review have been used to assess patient safety. One method for identifying adverse events is the Agency for Healthcare Research and Quality Patient Safety Indicators (AHRQ PSIs) which use hospital discharge data to report patient safety outcomes.1 Research using these indicators suggests that adverse events during hospital admissions of adults and children lead to excess length of stay (LOS) and increased hospital charges and mortality.2,3,4 The safety indicators system focuses on assessing the quality of care during a given inpatient hospitalization. Accordingly, it attempts to exclude patient healthcare injuries that occurred before the index hospital admission, such as during outpatient care or a previous admission. It also focuses on a relatively limited number of indicators designed to reflect a high probability of poor quality care during that admission.
In contrast, major studies based on chart review have attempted to achieve a more comprehensive assessment of patient safety problems by including the full spectrum of adverse events and, specifically, those caused by medical care before as well as during the index hospital admission.5 An alternative assessment method uses the Wisconsin Medical Injury Prevention Program (WMIPP) screening criteria. The WMIPP screening criteria identify medical injuries in routinely collected hospital discharge data as part of a confidential preventive model of hospital patient safety.6 Compared with medical record review, the adjusted sensitivity of the screening criteria is 60% and the specificity is 97%.7 WMIPP criteria include some specific additional diagnosis codes not found in the AHRQ PSIs in the areas of infection during medical care, transfusion reaction, and anesthesia complications. The AHRQ PSIs focus on potential complications and adverse events to identify 20 types of inpatient medical injuries while our criteria include health care provided in inpatient, ambulatory, community, and home settings that are associated with a medical injury identified during a subsequent or concomitant hospital admission. For example, an adolescent may ingest an overdose of acetaminophen at home with subsequent admission to hospital and a discharge diagnosis of acetaminophen poisoning. A hospital admission suggests the severity of the medical injury, while the occurrence at home suggests the need for community based prevention.
This study examined the incidence and outcomes of medical injuries among children admitted to hospital in Wisconsin using the WMIPP screening criteria. The primary research question was to determine the medical injury rates in four categories (medications; surgical and medical procedures; devices, implants and grafts; and radiation) and 40 subcategories. The secondary research question was to compare the length of stay and total hospital charges for children with and without medical injuries after adjusting for the All Patient Refined‐Diagnosis Related Groups (APR‐DRG) disease category, illness severity, mortality risk, and clustering within hospitals. We hypothesized that our findings would be consistent with previous published reports focused on inpatient errors and indicators, but would broaden the understanding of pediatric patient safety priorities in the context of a public health injury prevention model.
Methods
Data source
The research design was a cross sectional review of hospital discharge records for children aged 0–17 years from the State of Wisconsin Bureau of Health Information for the period from 1 January 2000 to 31 December 2002. The State mandates reporting from all general, acute care, non‐federal hospitals in Wisconsin. Like many state agencies or hospital associations, the Bureau routinely collects, edits, and publicly disseminates summary data based on the Health Care Finance Administration uniform billing report form (UB‐92) on all patients discharged from hospitals in Wisconsin. The UB‐92 data include International Classification of Disease (ICD‐9‐CM) N‐codes and, where relevant, E‐codes. In cases of injury, E‐codes attribute the injury to an external cause such as therapeutic misadventure, whereas N‐codes describe the nature of the injury such as a complication particular to a specific surgical procedure, but do not attempt to attribute the cause of the injury. The Bureau cleans the data set so all records are complete and required data are reported and consistent.
WMIPP medical injury screening
Medical injuries were classified into four broad categories: medications; surgical and medical procedures; devices, implants and grafts; and radiation. These categories were further divided into 40 subcategories that indicate more precisely the cause of the injury. For example, subgroups of medication injuries include complications due to poisoning, accidental poisoning, and adverse reaction with proper administration. Similarly, the subgroups of device, implant or graft injuries include infection, inflammation, mechanical and other complications. The classification scheme and the ICD‐9‐CM criteria for each category and subcategory are listed in the online Appendix available at http://www.qshc.com/supplemental.
Hospital characteristics were obtained from the 2001 Wisconsin Bureau of Health Information Annual Survey of Hospitals. The Medical College of Wisconsin institutional review board approved this study involving human subject records.
Data analysis
Rates of medical injuries were calculated as the number of children discharged with the particular type of medical injury divided by the total number of discharges of children admitted to hospital in Wisconsin during the study period. Statistical analysis was performed using Stata software release 8.0 (StataCorp, College Station, TX, 2003).
We calculated the impact of medical injury on LOS and total hospital charges using linear regression on log transformations of the LOS and charge data for patient discharges with and without medical injury. For analytical purposes we added 1 to the routinely reported LOS so that the reported LOS measure indicated the number of days on which the patient was hospitalized. As a result, patients admitted and discharged on the same day would be considered to have an LOS of 1 day. We determined mean excess LOS and percentage increase in LOS compared with discharges without the specified medical injury using a log transformation of LOS to normalize the variable and after adjusting for the All Patient Refined‐Diagnosis Related Groups (APR‐DRG) disease classification version 15.0 (3M Health Information Systems, 1998) and indices for risk of mortality and severity of illness calculated by the APR‐DRG system. Because the indices for risk of mortality and severity of illness were intended to adjust for the severity of the underlying illness in patients with and without a medical injury, they were calculated after excluding all medical injury related diagnostic codes. To account for within hospital similarities or clustering we used the Huber/White/sandwich estimator of variance in the regression model.
We could not assign a severity of underlying illness score to 970 patient records (0.3%), primarily because, after exclusion of the medical injury codes, no codes valid as a principal diagnosis remained. Patients who could not be assigned a severity of illness score were excluded from the analyses of LOS. Patients who could not be assigned a severity of illness score, who had a principal diagnosis of “V650” (healthy person accompanying sick person), or who had no hospital charges were excluded from the analyses of hospital charge data.
Results
We studied 318 785 discharge records from 134 general, acute care, non‐federal hospitals serving children in Wisconsin from January 2000 through December 2002 (table 1). The proportion of all discharges that were children at these hospitals varied widely; range 0.5–99%, mean (SD) 16 (15)%. In four hospitals over 50% of all the discharges were children. Most of the hospitals were non‐profit organizations without residency training programs or trauma centers, while 24% had residency programs and 27% were community trauma centers.
Table 1 Characteristics of Wisconsin hospitals, 2000–2.
| No (%) of hospitals* | |
|---|---|
| Ownership | |
| Non‐profit | 86 (69%) |
| Religious | 35 (28%) |
| County | 2 (2%) |
| For profit | 4 (3%) |
| Residencies | |
| None | 95 (76%) |
| 1–2 | 21 (17%) |
| 3+ | 9 (7%) |
| Self‐designated trauma level | |
| None | 67 (54%) |
| Rural center | 20 (16%) |
| Community center | 34 (27%) |
| Regional resource | 4 (3%) |
| Pediatric proportion of all discharges (mean (SD) %) | 16 (15) |
| Total | 125 (100%) |
*Although discharge records from 134 hospitals were analyzed, nine of these hospitals did not have survey information about characteristics due to the hospital closing before the 2001 survey (n = 2) or opening subsequent to the survey (n = 7).
Sixty four percent of the pediatric patients were normal full term newborns discharged within 7 days of age; 12% were adolescents (table 2). Slightly more than 50% were male. Most were privately insured, 25% were enrolled in Medicaid or the State Child Health Insurance Program, and 3% were uninsured; 74% had the lowest severity of illness classification and 94% had the lowest mortality risk; 92% were medical patients; 95% had routine discharges to home. The overall mean LOS was 4 days but 4% had stays of 10 days or longer.
Table 2 Characteristics of pediatric patients in Wisconsin hospitals, 2000–2.
| No (%) pediatric patients | |
|---|---|
| Age | |
| <8 days | 205584 (64%) |
| 8–365 days | 22953 (7%) |
| 1–4 years | 25446 (8%) |
| 5–12 years | 27134 (9%) |
| 13–17 years | 37668 (12%) |
| Sex | |
| Male | 162850 (51%) |
| Female | 155935 (49%) |
| Primary payer | |
| Private | 227578 (71%) |
| Public | 81805 (26%) |
| Self‐pay | 9402 (3%) |
| Condition on admission | |
| Normal term newborns | 189096 (59%) |
| Sick or premature newborns | 6536 (2%) |
| Extramural birth | 3759 (1%) |
| Urgent | 67987 (21%) |
| Emergency | 30446 (10%) |
| Elective | 20961 (7%) |
| APR‐DRG severity class | |
| 1 (lowest) | 236523 (74%) |
| 2 | 58351 (18%) |
| 3 | 18037 (6%) |
| 4 (highest) | 4901 (2%) |
| APR‐DRG mortality risk | |
| 1 (lowest) | 297807 (94%) |
| 2 | 13242 (4%) |
| 3 | 4990 (2%) |
| 4 (highest) | 1773 (1%) |
| Medical or Surgical APR‐DRG | |
| Medical | 292536 (92%) |
| Surgical | 26249 (8%) |
| Discharge status* | |
| Routine discharge | 302068 (95%) |
| Home under care | 7747 (2%) |
| Transfer to other hospital or facility | 7657(2%) |
| Expired | 1042 (<1%) |
| Left against medical advice | 237 (<1%) |
| Hospice care | 34 (<1%) |
| Overall mean (SD) LOS (days) | 4.3 (7.1) |
| Total | 318785 |
APR‐DRG, All Patient Refined‐Diagnosis Related Groups; LOS, length of stay
*One hospital incorrectly reported discharge status during 2000–1. While this error cannot be corrected, we estimate that only routine discharge and home under care categories were affected sufficiently to alter their distribution. Their true proportions are estimated to be 94% and 3%, respectively.
The WMIPP screening criteria identified 10 850 pediatric discharges (3.4% of all discharges) with 12 038 medical injuries (table 3). Medical injuries were attributed to medications in 41%; surgical and medical procedures in 35%; devices, implants and grafts in 24%; and radiation in <1%. The rate of medical injury among all pediatric discharges was 1.5% for medications, 1.3% for surgical and medical procedures, 0.9% for devices, implants, and grafts, and 0.01% for radiation. The highest rates were found with non‐specific device complications, non‐narcotic analgesic/antipyretic/anti‐rheumatic drugs (such as acetaminophen overdoses), unclassified procedure complications, vascular devices, systemic drugs, respiratory complications of procedures, other psychotropic medications, antibiotic complications, and hematomas/hemorrhage or infections/inflammation after a procedure.
Table 3 Medical injuries and related outcomes for pediatric patients in Wisconsin hospitals, 2000–2.
| Categories (and subcategories) of WMIPP screening criteria | No of cases | Rate per 10 000 discharges | Adjusted mean excess LOS (days)‡ | Adjusted % change in LOS (days)‡ | Adjusted mean excess hospital charges ($)‡ |
|---|---|---|---|---|---|
| Complications due to medications | 4918 | 154.8 | 0.0 | 1.1 | 1437** |
| Antibiotics | 530 | 16.7 | 0.9** | 22.1** | 2019** |
| Hormones | 279 | 8.8 | 0.6** | 13.8** | 2255** |
| Systemic agents | 879 | 27.7 | 0.2 | 5.2 | 1249** |
| Agents affecting blood constituents | 40 | 1.3 | 0.1 | 1.3 | 1945** |
| Blood products | 150 | 4.7 | 0.5 | 12.6 | 1386** |
| Non‐narcotic analgesics, antipyretics, anti‐rheumatics | 1184 | 37.3 | −0.5** | −12.7** | 1844** |
| Narcotic analgesics | 438 | 13.8 | 0.5** | 12.0** | 1796** |
| Anticonvulsants, anti‐Parkinson drugs | 269 | 8.5 | −0.3 | −7.3 | 553 |
| Sedatives, hypnotics | 320 | 10.1 | −0.1 | −2.4 | 1230** |
| Other psychotropics | 613 | 19.3 | −1.0** | −24.0** | 68 |
| Autonomic nervous system | 192 | 6.0 | −0.3 | −5.9 | 1237** |
| Cardiovascular drugs | 104 | 3.3 | −0.8* | −18.4* | −404 |
| Gastrointestinal, smooth muscle, and respiratory drugs | 121 | 3.8 | −0.9** | −20.9** | −112 |
| Water, mineral, and uric acid drugs | 85 | 2.7 | 1.4** | 33.8** | 4064** |
| Anesthesia | 77 | 2.4 | −0.3 | −6.5 | 553 |
| CNS stimulants, topical drugs, vaccines, other miscellaneous drugs | 216 | 6.8 | −0.9** | −20.6** | −118 |
| Specific reaction to unknown drug | 418 | 13.2 | 1.0** | 22.4** | 1497** |
| Miscellaneous complications due to unspecified drugs | 476 | 15.0 | −0.5** | −12.8** | 1050** |
| Surgical and medical procedures | 4194 | 132.0 | 0.9** | 21.8** | 1333** |
| Complication of amputation or other removed organ | 105 | 3.3 | 0.8** | 18.7** | 628 |
| Tracheostomy | 159 | 5.0 | 0.9** | 20.5** | 1920* |
| Formation of a stoma | 466 | 14.7 | 1.0** | 24.1** | 2007** |
| Gastrointestinal complication | 503 | 15.8 | 1.3** | 29.7** | 1311** |
| Cardiac complication | 262 | 8.2 | 0.8** | 19.2** | 1119** |
| Vascular complication | 193 | 6.1 | 1.0** | 22.9** | 1597** |
| Respiratory complication | 722 | 22.7 | 0.4** | 10.5** | 466* |
| Nervous system complication | 151 | 4.8 | 1.5** | 34.0** | 1355** |
| Genitourinary complication | 117 | 3.7 | 1.0** | 22.2** | 764* |
| Infection, inflammation | 503 | 15.8 | 2.0** | 47.1** | 2398** |
| Hematoma, hemorrhage | 517 | 16.3 | 0.2 | 5.8 | 416 |
| Non‐healing wound | 218 | 6.9 | 1.7** | 40.7** | 1895** |
| Miscellaneous complications to specific procedures | 41 | 1.3 | 1.0* | 23.7* | 1610 |
| Misadventure | 256 | 8.1 | 0.8** | 18.8** | 1280** |
| Complications not classified elsewhere | 1168 | 36.8 | 0.8** | 17.7** | 1403** |
| Devices, implants and grafts | 2880 | 90.6 | 1.1** | 26.8** | 2508** |
| Cardiac | 88 | 2.8 | 0.6** | 14.0** | 3051** |
| Vascular | 946 | 29.8 | 2.1** | 49.5** | 4465** |
| Orthopedic | 190 | 6.0 | 0.6** | 13.0** | 1429** |
| Genitourinary | 70 | 2.2 | 2.4** | 56.1** | 3803** |
| Transplanted organ or body part | 463 | 14.6 | 1.3** | 29.2** | 3726** |
| Other complications | 1468 | 46.2 | 0.5** | 12.1** | 1118** |
| Radiation | 46 | 1.5 | −0.1 | −1.6 | 1032* |
| All radiation related injuries | 46 | 1.5 | −0.1 | −1.6 | 1032* |
| Total | 10850† | 3.41 | 0.51** | 11.9** | 1614** |
WMIPP, Wisconsin Medical Injury Prevention Program; LOS, length of stay.
*p<0.05, **p<0.01
†The number of injured patients in each category adds to more than the total number of injured patients because some patients had a medical injury in more than one category. 10 850 patients had a total of 12 038 medical injuries.
‡Excess length of stay (LOS), percentage change in LOS, and total hospital charges were adjusted for APR‐DRG disease category, illness severity, mortality risk, and clustering within hospitals.
The mean LOS adjusted for APR‐DRG disease category, illness severity, mortality risk and clustering within hospitals was 4.8 days for patients with medical injuries and 4.3 days for patients without medical injuries (table 4). This half day difference represents a statistically significant 12% excess mean LOS associated with medical injuries. The regression model results for LOS and hospital charges are listed in table 5.
Table 4 Adjusted LOS and hospital charges for Wisconsin pediatric patients with and without medical injuries, 2000–2.
| Adjusted mean LOS (days) | Adjusted mean total hospital charges ($) | |
|---|---|---|
| Patients with medical injury | 4.78 (n = 10 850) | 7774 (n = 10 850) |
| Patients without medical injury | 4.27 (n = 306 962) | 6160 (n = 306 894)* |
| Absolute difference | 0.51 | 1614 |
| Percentage difference (95% CI) | 11.9% (6.9 to 17.1) | 26.2% (22.1 to 30.5) |
LOS, length of stay; CI, confidence interval.
Mean LOS and total hospital charges adjusted for APR‐DRG disease category, illness severity, mortality risk, and clustering within hospitals.
*68 records were excluded because they had missing charge data. 58 of these cases were expired newborns, four were instances of a healthy person accompanying a sick person, and three were stays over 100 days for which charge data are not required. It is unknown why the remaining two cases had no charges.
Table 5 Regression model results for length of stay and hospital charges.
| Adjusted covariates in regression models | Increased LOS % (95% CI ) | Increased hospital charges % (95% CI) |
|---|---|---|
| Medical injury | 11.9 (6.9 to 17.1)* | 26.2 (22.1 to 30.5)* |
| Severity level | ||
| 1 (lowest) | Reference group | Reference group |
| 2 | 20.8 (18.5 to 23.1)* | 47.4 (42.0 to 53.0)* |
| 3 | 62.0 (56.6 to 67.5)* | 247.9 (229.1 to 268.2)* |
| 4 | 325.5 (272.1 to 389.5)* | 612.8 (458.9 to 818.2)* |
| Risk of mortality level | ||
| 1 (lowest) | Reference group | Reference group |
| 2 | 5.1 (1.6 to 8.8)* | 8.6 (2.2 to 15.5)* |
| 3 | 3.7 (−8.6 to 17.5) | 18.3 (1.5 to 37.9)* |
| 4 | −25.3 (−49.6 to 10.7) | −2.0 (−41.6 to 64.6) |
*p<0.05.
APR‐DRG was also a covariate in the two models but the categories are too numerous to list.
The similarly adjusted mean total hospital charges were $7774 for patients with medical injuries and $6160 for patients without medical injuries. This difference of $1614 represents a 26% increase in the mean charge associated with medical injuries, which is statistically significant. The adjusted mean excess LOS and percentage change associated with medical injuries was highest with genitourinary devices/implants, vascular devices, infections/inflammation and non‐healing wounds after procedures, water/mineral/uric acid drugs, gastrointestinal complications due to a procedure, specific reactions to unknown drugs, and antibiotic complications (table 3). Medical injuries with adjusted mean excess hospital charges exceeding $2000 were associated with vascular devices (mean increased charge $4465), water/mineral/uric acid drugs, genitourinary devices/implants, transplanted organs/body parts, infections/inflammation after procedures, hormone medications, antibiotic complications, and stoma formation after procedures. Although medical injuries associated with a few selected medication subcategories (such as other psychotropics, gastrointestinal/smooth muscle/respiratory, and central nervous system stimulants/topicals/vaccines/miscellaneous) had statistically significant decreased adjusted mean LOS, the decrease in adjusted mean hospital charges was not significant and less than $120. By contrast, radiation injuries were associated with no significant change in adjusted mean LOS but a statistically significant increase in adjusted mean charges in excess of $1000.
Discussion
This is the first report of medical injury rates among children in Wisconsin hospitals and of hospital safety outcomes analyzed through 2002 for a large population of children. The study uses the new WMIPP screening criteria as a comprehensive surveillance tool to assess hospital measures of child health and safety. The criteria identify medical injuries that may occur in inpatient, ambulatory, home, or other community settings and are diagnosed in hospital settings. Our focus on medical injuries requires both a clinical and a public health approach to preventing these adverse events. Healthcare professionals can use medical injury surveillance to guide clinical quality improvement initiatives. Public health leaders can use this information to prioritize areas for outpatient or community interventions—for example, healthcare professionals, community representatives, government officials, and family members might combine to develop preventive strategies to reduce hospital admissions for acetaminophen overdoses among adolescents.
Our results generally are consistent with AHRQ PSI findings published by Miller et al.3 In their analysis of 3.8 million pediatric discharge records from 22 states in 1997, AHRQ PSI events were associated with 2–6‐fold longer LOS and 2–20‐fold higher total charges. In our study specific medical injuries had up to 56% excess adjusted mean LOS and $4465 excess mean adjusted hospital charges. Multivariate analyses by Miller et al showed that all AHRQ PSI events except birth trauma were directly associated with factors related to greater illness severity—that is, PSI events occurred more frequently in those with more severe illness.3 In a second AHRQ PSI analysis of 7.45 million hospital discharges of patients of all ages from 28 states in 2000, postoperative sepsis and wound dehiscence were the most serious events in terms of extra LOS and excess hospital charges.2 In our study extra LOS and excess hospital charges were greater for injuries due to genitourinary devices/implants, vascular devices, and infections/inflammation after procedures. In a third AHRQ PSI analysis of 5.7 million pediatric discharge records from 27 states in 2000, AHRQ PSI events occurred more frequently in the very young and those enrolled in Medicaid.4 WMIPP criteria do not capture any of the AHRQ PSI birth trauma codes. We did not analyze medical injuries by age, payer, or other patient characteristics because this study focused on hospital outcomes rather than patient demographic characteristics. Furthermore, the State of Wisconsin does not report racial/ethnic background of patients in publicly available hospital records. Accordingly, we cannot interpret our findings along these parameters.
The AHRQ PSIs focus exclusively on hospital inpatient safety problems. As a result of their focus on inpatient quality of care, the AHRQ PSIs are more appropriate tools to evaluate the quality of care in hospitals. However, our approach provides a more accurate estimate of the full scope of patient safety problems as it includes medical injuries identified or treated in a hospital even if injury took place in the outpatient, community, or home setting, or during a previous hospital admission. In addition, this approach can target certain patient safety problems such as drug reactions and device failures which frequently occur in an outpatient setting.
Our study and the work of others show that medication related injuries are an important and common problem in pediatrics. In our study, medical injuries due to medications were diagnosed in 1.5% of hospitalized children and were associated with an excess mean hospital charge of $1437. In a cohort study of hospitalized non‐newborn pediatric patients in the US in 1988, 1991, 1994, and 1997, the rate of hospital reported medical errors ranged from 1.8% to 3.0%. Children with special needs or dependence on medical technology experienced significantly higher rates of medical errors.8 In a prospective study of 1197 pediatric admissions at one center, adverse drug events occurred in 6% and potential events in 8% of hospitalized children, especially in those with a greater disease burden and medication exposure.9 Use of computerized physician order entry with clinical decision support systems, ward based clinical pharmacists, and improved communication among physicians, nurses, and pharmacists might substantially reduce medication error rates and prevent potential adverse drug events in pediatric inpatients.10,11,12,13 The magnitude of benefits may be even greater in pediatric than in adult medicine because of the need for weight based dosing.
This approach to patient safety monitoring has limitations. It is based on hospital discharge data collected for administrative purposes and shares the limitations of all studies using administrative data.14 Because they are collected primarily for administrative purposes, hospital discharge data cannot include deep clinical detail. Due to privacy concerns, these data also do not have patient identifiers, so multiple admissions for the same person cannot be linked. The Wisconsin public use data set also lacks important information about patient race/ethnicity. In this study medical injuries cannot be attributed to a hospital or to inpatient care due to inter‐hospital transfers and injuries occurring in outpatient and non‐medical settings. At the same time, administrative data have advantages of being readily available, inexpensive, computer readable, and covering large populations.15 They are therefore useful for medical injury surveillance but efforts need to be made to overcome selected limitations.
The use of administrative indicators as useful valid outcome measures has limitations including problems with variations in coding accuracy and practice, challenges to appropriate risk adjustment, and difficulty with correct attribution of the timing of events.16,17,18,19 The sensitivity of the WMIPP screening criteria is approximately 60%. Given the relatively low statistical prevalence of medical injuries, a positive screening has a predictive value of 50–60%. The descriptive epidemiology of injuries described by our WMIPP method therefore only has a 50–60% chance of being accurate. This may not be reassuring even if, on average, those with positive screens have higher costs and LOS for some categories and overall. The same criticism can be made of the AHRQ PSIs but it has not slowed their adoption.
While the WMIPP method provides a more public health oriented approach to medical injury than PSI, it is consistent with other administrative data based tools in that it does not separate the injuries into pre‐hospital or in‐hospital phases, nor does it indicate which are injuries as a result of ambulatory care or even intentional poisoning such as during a suicide attempt. Although the finding that the most significant injuries relate to devices and procedures—most likely inpatient related—makes this distinction less relevant, it also lowers the relative value of using the WMIPP measures compared with the PSI.
These limitations serve as a caution to public reporting for accountability, but do not preclude the use of medical injury screening of administrative data for improving patient care. The WMIPP criteria reflect the public health impact of medical injuries better than systems focused on inpatient events only. Medical injuries that are challenging to prevent with current knowledge and technology may be preventable with future greater knowledge and new technology, medications and devices.
Our approach is also limited by the variability among states in external cause of injury coding in hospital discharge data.20 In Wisconsin, external cause of injury coding is very complete.21 In this pediatric dataset, E‐coding is 80% complete. Although our approach uses N‐codes as well as E‐codes, WMIPP screening criteria would be of greatest use in this and other states with high levels of E‐coding. Despite these limitations, the face validity of the diagnostic codes comprising the criteria, the agreement of the criteria with medical record review,7 and the adverse effect on LOS and charges experienced by individuals fulfilling the criteria all indicate that the WMIPP criteria are useful in identifying injuries from healthcare interventions that have an appreciable impact on patient health.
We think that actual specific medical injuries are under‐reported by Wisconsin hospitals. Patients may under‐report symptoms and physicians may under‐report specific diagnoses associated with medical injuries—for example, therapeutic misadventures and complications associated with childbirth. Moreover, non‐specific and unspecified codes, as defined in the ICD‐9‐CM manual, may be over‐represented as they account for 10.8% and 12.3% of all codes reported in this dataset. Individual hospitals had an average of 10.1% non‐specific codes (range 0–50%) and 14.7% unspecified codes (range 3.3–50%).
Key messages
Inpatient medical injuries are common and result in excess length of stay and increased hospital charges.
This study applied new screening criteria that included healthcare injuries occurring before rather than just during a hospital admission.
The results reinforce previous national findings to the end of 2000 using Wisconsin data to the end of 2002.
They suggest that hospitals and pediatricians should use more specific diagnostic coding, focus clinical improvement on medication, procedures and devices frequently associated with medical injuries, and use medical injury surveillance to track medical injury rates in children.
Medical injuries may occur with or without an error in medical practice. The issue of whether medical injuries are caused by errors is complicated by variations in definitions of errors and lack of reproducibility in error determination. Cook and Woods22 have identified multiple reasons for the limitations of focusing safety efforts on error including ease of identifying a human proximal to a failure as a sole cause, the difficulty in identifying the “causal chain that led to the system failure”, the usual success of human performance in the same flawed and complex systems and, lastly, hindsight bias. Additionally, focusing on error leads to attributions of blame23 which does little to advance injury prevention.
Our approach and the results of this study have important generalizable implications for the practice of pediatrics, healthcare improvements, and the betterment of child and adolescent health.24 The application of the WMIPP screening criteria to large population based databases permits public health surveillance of patient safety.
The classification scheme and ICD‐9‐CM criteria for each category and subcategory are listed in the online Appendix available at http://www.qshc.com/supplemental.
Supplementary Material
Acknowledgements
The authors acknowledge the valuable efforts of the Wisconsin Medical Injury Prevention Program Research Group including Chris McLaughlin, Linda N Meurer, Michele Leininger, Jean Grube, Karen J Brasel, Stephen W Hargarten, Janice B Babcock, Evelyn M Kuhn, and Prakash Laud, and thank Emmanuel Ngui for reviewing the manuscript.
Abbreviations
AHRQ - Agency for Healthcare Research and Quality
APR‐DRG - All Patient Refined‐Diagnosis Related Groups
LOS - length of stay
PSI - patient safety indicator
WMIPP - Wisconsin Medical Injury Prevention Program
Footnotes
This research project was funded by the Agency for Healthcare Research and Quality grant U18‐HS11893 to the Medical College of Wisconsin. 3M Health Information Systems granted the All Patient Refined‐Diagnosis Related Groups software license. This work was partially supported by the Centers for Disease Control and Prevention grant R49/CCR519614.
Competing interests: none.
The authors are solely responsible for the design, conduct, and analysis of the study and development of this manuscript.
The classification scheme and ICD‐9‐CM criteria for each category and subcategory are listed in the online Appendix available at http://www.qshc.com/supplemental.
References
- 1.Miller M R, Elixhauser A, Zhan C.et al Patient safety indictors: using administrative data to identify potential patient safety concerns. Health Serv Res 200136110–132. [PMC free article] [PubMed] [Google Scholar]
- 2.Zhan C, Miller M R. Excess length of stay, charges, and mortality attributable to medical injuries during hospitalization. JAMA 20032901868–1874. [DOI] [PubMed] [Google Scholar]
- 3.Miller M R, Elixhauser A, Zhan C. Patient safety events during pediatric hospitalizations. Pediatrics 20031111358–1366. [DOI] [PubMed] [Google Scholar]
- 4.Miller M R, Zhan C. Pediatric patient safety in hospitals: a national picture in 2000. Pediatrics 20041131741–1746. [DOI] [PubMed] [Google Scholar]
- 5.Brennan T A, Leape L L, Laird N M.et al Incidence of adverse events and negligence in hospitalized patients results: results of the Harvard Medical Practice Study I. N Engl J Med 1991324370–376. [DOI] [PubMed] [Google Scholar]
- 6.Layde P M, Cortes L M, Teret S P.et al Patient safety efforts should focus on medical injuries. JAMA 20022871993–1997. [DOI] [PubMed] [Google Scholar]
- 7.Layde P M, Meurer L N, Guse C.et al Medical injury identification using hospital discharge data. In: Henricksen K, Battles JB, Marks E, Lewin DI, eds. Advances in patient safety: from research to implementation. Rockville, MD: Agency for Healthcare Research and Quality, 20052119–132. [PubMed] [Google Scholar]
- 8.Slonim A D, LaFleur B J, Ahmed W.et al Hospital‐reported medical errors in children. Pediatrics 2003111617–621. [DOI] [PubMed] [Google Scholar]
- 9.Holdsworth M T, Fichtl R E, Behta M.et al Incidence and impact of adverse drug events in pediatric inpatients. Arch Pediatr Adolesc Med 200315760–65. [DOI] [PubMed] [Google Scholar]
- 10.Kaushal R, Bates D W, Landrigan C.et al Medication errors and adverse drug events in pediatric inpatients. JAMA 20012852114–2120. [DOI] [PubMed] [Google Scholar]
- 11.Kaushal R, Barker K N, Bates D W. How can information technology improve patient safety and reduce medication errors in children's health care? Arch Pediatr Adolesc Med 20011551002–1007. [DOI] [PubMed] [Google Scholar]
- 12.Kaushal R, Shojania K G, Bates D W. Effects of computerized physician order entry and clinical decision support systems on medication safety: a systematic review. Arch Intern Med 20031631409–1416. [DOI] [PubMed] [Google Scholar]
- 13.Fortescue E B, Kaushal R, Landrigan C P.et al Prioritizing strategies for preventing medication errors and adverse drug events in pediatric inpatients. Pediatrics 2003111722–729. [DOI] [PubMed] [Google Scholar]
- 14.Hunt J P, Baker C C, Fakhry S M.et al Accuracy of administrative data in trauma. Surgery 1999126191–197. [PubMed] [Google Scholar]
- 15.Zhan C, Miller M R. Administrative data‐based patient safety research: a critical review. Qual Saf Health Care 200312(Suppl II)ii58–ii63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Schwartz R M, Gagnon D E, Muri J H.et al Administrative data for quality improvement. Pediatrics 1999103291–301. [PubMed] [Google Scholar]
- 17.Weiss J, Nannini A, Fogerty S.et al Use of hospital discharge data to monitor uterine rupture—Massachusetts, 1990–1997. MMWR 200049245–248. [PubMed] [Google Scholar]
- 18.Iezzoni L I. Assessing quality using administrative data. Ann Intern Med 1997127666–674. [DOI] [PubMed] [Google Scholar]
- 19.Romano P S, Mark D H. Bias in the coding of hospital discharge data and its implications for quality assessment. Med Care 19943281–90. [DOI] [PubMed] [Google Scholar]
- 20.Centers for Disease Control and Prevention External cause‐of‐injury coding in hospital discharge data – United States MMWR199443465–467. [PubMed] [Google Scholar]
- 21.Tavris D R.et al Hospitalizations for vehicle associated injuries in Wisconsin. Wisc Med J 19999834–39. [PubMed] [Google Scholar]
- 22.Cook R I, Woods D D. Operating at the sharp end: the complexity of human error. In: Bogner MS, ed. Human errors in medicine. Hillsdale, NJ: Erlbaum, 1994
- 23.Heider F.The psychology of interpersonal relationships. New York: Wiley, 1958
- 24.Meurer J R, Meurer L N, Grube J.et al Combining performance feedback and evidence‐based educational resources. In: Henricksen K, Battles JB, Marks E, Lewin DI, eds. Advances in patient safety: from research to implementation. Rockville, MD: Agency for Healthcare Research and Quality, 20054237–252. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.