Diagnostic Errors in the Pediatric and Neonatal ICU
Diagnostic Errors in the Pediatric and Neonatal ICU
This systematic review builds on recent data describing diagnostic errors in the adult ICU and is the first to aggregate available data on PICU and NICU misdiagnosis. We found that 6.4% of autopsied PICU patients and 3.7% of autopsied NICU patients had a class I diagnostic error. We also found that 13.8% of autopsied PICU patients and 15.5% of autopsied NICU patients had a class II diagnostic error. Error rates in the PICU population are similar to recently published data on adult ICU patients, but the reported NICU data show a slightly lower overall error rate.
These results must be interpreted in the context of knowing that not all patients who died underwent autopsy. There is likely bias in selecting which patients undergo an autopsy (i.e., autopsies are thought to be more aggressively pursued when the cause of death is clinically uncertain), so measured error rates vary as a function of the autopsy rate and may differ from the "true" error rate where all patients were autopsied. We attempted to create a model-based estimate of the "true" class I error rate, but the data were too limited to make an accurate assessment. It would be difficult to estimate the burden of misdiagnosis on pediatric mortality with the available data. Adult data suggest that as many as 40,500 adult ICU patients per year die as a result of diagnostic error. Since our data show similar error rates to adults, the pediatric impact is likely to be significant. Autopsy-based studies cannot assess morbidity attributable to diagnostic error in patients who survive their ICU stay, but the impact on patients and the healthcare system is likely substantial. Diagnostic error is known to contribute to reoperations and increased length of stay in ICUs. Studies based on closed malpractice claims suggest that misdiagnosis results in serious morbidity as often as death.
Major errors (class I/II) most often involved missed infections (30.8%) and vascular events (18.3%). Overall, this is similar to adult ICU patients, but in adults, vascular events outnumber infections, whereas the reverse is true for children and neonates. Undiagnosed congenital malformations and heritable genetic/metabolic conditions also contributed to major diagnostic errors (16.2%). These were more often seen in NICU patients than in PICU patients and were much more likely to be considered lethal (i.e., class I). This is not surprising since PICUs typically treat older children in whom congenital disorders have a tendency to have been previously diagnosed.
There are limited data to determine the root cause of misdiagnosis and possible solutions. We were unable to draw broad inferences across studies because risk factors were inconsistently reported. The diagnostic process is complex and cognitive or system errors may occur at any step. One of the studies suggested that teams were at least aware of diagnostic uncertainty, contrary to what has been reported for adults. Information overload and poor data management have been suggested as possible reasons to explain why ICUs may have a higher prevalence of misdiagnosis than the general hospital population. Information technology that allows for more efficient retrieval and organization of patient data may help reduce ICU diagnostic error.
Pediatric hospitalized patients present special cognitive challenges because of the wide range of ages, developmental stages, and diagnoses cared for simultaneously. Unlike adults, age-specific diseases with variable presentations add to the diagnostic complexity. This is especially so for congenital malformations and genetic/metabolic disease. Advances in genetic testing may improve future diagnostic accuracy. Antemortem testing should be considered, although costs may also factor into the equation. Postmortem genetic and metabolic testing is becoming more frequent and may also be appropriate for pediatric patients where the cause of death is in question. Studies have shown that genetic and metabolic postmortem testing can identify a previously missed diagnosis approximately 18% of the time. Findings from postmortem genetic studies may have a significant impact on other family members as well as future offspring of the decedent's parents.
Numerous strategies to improve diagnostic accuracy have been proposed, but none has been tested for their impact on hard clinical outcomes. A targeted approach to diagnostic education in medical training may improve diagnostic reasoning skills. Enhanced training in the cognitive aspects of diagnosis and awareness of potential biases might help reduce cognitive error. Systems-oriented solutions to reduce diagnostic errors may hold special promise. Maintaining adequate staffing models, appropriate physician to patient ratios, and 24/7 presence of or access to expert clinicians in ICUs may reduce error rates. Computer-based diagnostic decision support tools are expected to transform medical diagnosis in the future and have been shown to be effective in some limited settings. Few tools, however, have been created for or tested in the complex PICU or NICU environments.
Discussion
This systematic review builds on recent data describing diagnostic errors in the adult ICU and is the first to aggregate available data on PICU and NICU misdiagnosis. We found that 6.4% of autopsied PICU patients and 3.7% of autopsied NICU patients had a class I diagnostic error. We also found that 13.8% of autopsied PICU patients and 15.5% of autopsied NICU patients had a class II diagnostic error. Error rates in the PICU population are similar to recently published data on adult ICU patients, but the reported NICU data show a slightly lower overall error rate.
These results must be interpreted in the context of knowing that not all patients who died underwent autopsy. There is likely bias in selecting which patients undergo an autopsy (i.e., autopsies are thought to be more aggressively pursued when the cause of death is clinically uncertain), so measured error rates vary as a function of the autopsy rate and may differ from the "true" error rate where all patients were autopsied. We attempted to create a model-based estimate of the "true" class I error rate, but the data were too limited to make an accurate assessment. It would be difficult to estimate the burden of misdiagnosis on pediatric mortality with the available data. Adult data suggest that as many as 40,500 adult ICU patients per year die as a result of diagnostic error. Since our data show similar error rates to adults, the pediatric impact is likely to be significant. Autopsy-based studies cannot assess morbidity attributable to diagnostic error in patients who survive their ICU stay, but the impact on patients and the healthcare system is likely substantial. Diagnostic error is known to contribute to reoperations and increased length of stay in ICUs. Studies based on closed malpractice claims suggest that misdiagnosis results in serious morbidity as often as death.
Major errors (class I/II) most often involved missed infections (30.8%) and vascular events (18.3%). Overall, this is similar to adult ICU patients, but in adults, vascular events outnumber infections, whereas the reverse is true for children and neonates. Undiagnosed congenital malformations and heritable genetic/metabolic conditions also contributed to major diagnostic errors (16.2%). These were more often seen in NICU patients than in PICU patients and were much more likely to be considered lethal (i.e., class I). This is not surprising since PICUs typically treat older children in whom congenital disorders have a tendency to have been previously diagnosed.
There are limited data to determine the root cause of misdiagnosis and possible solutions. We were unable to draw broad inferences across studies because risk factors were inconsistently reported. The diagnostic process is complex and cognitive or system errors may occur at any step. One of the studies suggested that teams were at least aware of diagnostic uncertainty, contrary to what has been reported for adults. Information overload and poor data management have been suggested as possible reasons to explain why ICUs may have a higher prevalence of misdiagnosis than the general hospital population. Information technology that allows for more efficient retrieval and organization of patient data may help reduce ICU diagnostic error.
Pediatric hospitalized patients present special cognitive challenges because of the wide range of ages, developmental stages, and diagnoses cared for simultaneously. Unlike adults, age-specific diseases with variable presentations add to the diagnostic complexity. This is especially so for congenital malformations and genetic/metabolic disease. Advances in genetic testing may improve future diagnostic accuracy. Antemortem testing should be considered, although costs may also factor into the equation. Postmortem genetic and metabolic testing is becoming more frequent and may also be appropriate for pediatric patients where the cause of death is in question. Studies have shown that genetic and metabolic postmortem testing can identify a previously missed diagnosis approximately 18% of the time. Findings from postmortem genetic studies may have a significant impact on other family members as well as future offspring of the decedent's parents.
Numerous strategies to improve diagnostic accuracy have been proposed, but none has been tested for their impact on hard clinical outcomes. A targeted approach to diagnostic education in medical training may improve diagnostic reasoning skills. Enhanced training in the cognitive aspects of diagnosis and awareness of potential biases might help reduce cognitive error. Systems-oriented solutions to reduce diagnostic errors may hold special promise. Maintaining adequate staffing models, appropriate physician to patient ratios, and 24/7 presence of or access to expert clinicians in ICUs may reduce error rates. Computer-based diagnostic decision support tools are expected to transform medical diagnosis in the future and have been shown to be effective in some limited settings. Few tools, however, have been created for or tested in the complex PICU or NICU environments.