Article Text
PDF
Abstract
Background Little is known about the incidence or significance of diagnostic error in the inpatient setting. We used a malpractice claims database to examine incidence, predictors and consequences of diagnosis-related paid malpractice claims in hospitalised patients.
Methods The US National Practitioner Database was used to identify paid malpractice claims occurring between 1 January 1999 and 31 December 2011. Patient and provider characteristics associated with paid claims were analysed using descriptive statistics. Differences between diagnosis-related paid claims and other paid claim types (eg, surgical, anaesthesia, medication) were assessed using Wilcoxon rank-sum and χ2 tests. Multivariable logistic regression was used to identify patient and provider factors associated with diagnosis-related paid claims. Trends for incidence of diagnosis-related paid claims and median annual payment were assessed using the Cochran-Armitage and non-parametric trend test.
Results 13 682 of 62 966 paid malpractice claims (22%) were diagnosis-related. Compared with other paid claim types, characteristics significantly associated with diagnosis-related paid claims were as follows: male patients, patient aged >50 years, provider aged <50 years and providers in the northeast region. Compared with other paid claim types, diagnosis-related paid claims were associated with 1.83 times more risk of disability (95% CI 1.75 to 1.91; p<0.001) and 2.33 times more risk of death (95% CI 2.23 to 2.43; p<0.001) than minor injury, after adjusting for patient and provider characteristics. Inpatient diagnostic error accounted for $5.7 billion in payments over the study period, and median diagnosis-related payments increased at a rate disproportionate to other types.
Conclusion Inpatient diagnosis-related malpractice payments are common and more often associated with disability and death than other claim types. Research focused on understanding and mitigating diagnostic errors in hospital settings is necessary.
- diagnostic errors
- patient safety
- hospital medicine
- medical error
- measurement/epidemiology
Statistics from Altmetric.com
Introduction
Diagnostic errors are common, costly and morbid; however, the contribution of diagnostic error to health outcomes and economic expense remains largely underappreciated.1 In its seminal report, Improving Diagnosis in Healthcare, the Institute of Medicine defined diagnostic error as ‘the failure to (a) establish an accurate and timely explanation of the patient’s health problem(s) or (b) communicate that explanation to the patient.’2 Additionally, the report outlined several reasons why diagnostic error has been overlooked, including sparse data related to errors, challenges in defining and measuring when errors occurred, and data limited to retrospective study designs.
In the outpatient setting, diagnostic error is estimated to occur in 5.1% of encounters (or approximately 12 million patients annually).3 A recent analysis of all paid malpractice claims in the US National Practitioner Data Bank (NPDB) between 1992 and 2014 found that diagnostic error was the most common type of allegation.4 In a study published in 2013, Saber Tehrani et al summarised 25 years of diagnostic error claims in ambulatory and hospital settings, and concluded that diagnostic errors were the most common, costly and dangerous type of medical mistake.5 Because the report combined all venues of care, the precise impact of diagnostic errors in inpatient settings was less clear. The Controlled Risk Insurance Company (a captive risk management company formed by the Harvard Medical Institutions) reviewed internal claims from 2008 to 2012 and concluded that 26.0% of all claims alleging diagnostic error occurred in the inpatient setting.6 Similarly, a study of over 7000 patient records in the Netherlands reported that errors in diagnosis accounted for 6.5% of all hospital adverse events.7 Furthermore, autopsy data of patients who died in the intensive care unit suggest that 6.3% of deaths occurred due to misdiagnosis and 28% demonstrated at least one missed diagnosis.8 Collectively, these data suggest that inpatient errors are costly, with mean per-claim payment being significantly higher in inpatient than outpatient settings.5
Despite the work that has already been done to understand the incidence, types and costs related to diagnostic error in general, inpatient diagnostic errors have not been studied in detail. Specific patient or provider factors that are associated with inpatient diagnostic errors are unknown. Therefore, we conducted a study to better understand the trends, costs, outcomes and predictors associated with paid diagnosis-related malpractice claims.
Methods
Study design and population
We performed a retrospective analysis of paid medical malpractice claims reported in the NPDB Public Use Data File from 1 January 1999 to 31 December 2011.9 Established in 1986, the NPDB is an electronic archive of medical malpractice claims paid on behalf of, and adverse actions (eg, loss of licensure) taken against, physicians (both doctor of medicine and doctor of osteopathic medicine). Claims are reported once they have been closed. Claims resolved solely on behalf of an institution are not included. Data on setting of malpractice allegation (inpatient or outpatient) and severity of injury were not collected prior to 1 January 2004; thus, claims closed prior to 2004 were excluded. Claims resulting in non-monetary penalties (both before and after 1 January 2004) failed to include setting of malpractice allegation and were also excluded. Prior studies utilising malpractice claims data have described an average of 4.7 years between claim filing and closure for diagnostic error-related claims.5 Therefore, claims filed from 1 January 1999, 5 years prior to the designation of malpractice setting, to 31 December 2011, roughly 5 years prior to the end of the data set were included in the analysis, as claim data before and after those dates, respectively, were considered incomplete. Payments occurring between 1 January 2004 and 30 September 2016 were included in the analysis.
Variables and definitions
Both patient and provider characteristics (as available in the NPDB) were included in the analysis. Patient variables were limited to age and gender while provider characteristics included age, field of license (allopathic vs osteopathic physician), training status (resident vs attending) and region of license (Northeast, Midwest, South, West and other). Claims were labelled as inpatient if they occurred in an inpatient setting or spanned both inpatient and outpatient settings; claims limited to the emergency department were not differentiated in the database. Adverse events resulting in paid claims were categorised based on gradations of severity in the NPDB and included ‘minor’, ‘disability’ and ‘death’. For the purposes of our analysis, emotional, insignificant, minor and major temporary, and minor permanent injuries were considered to be minor whereas significant or major permanent injury, paralysis and brain damage were considered disability. Physicians with multiple claims were anonymously identified using the unique practitioner identification number.
Raw payment data (defined as the total amount paid to a plaintiff on behalf of a practitioner in a single case) were coded into ranges with average of the maximum and the minimum value in the range calculated per NPDB recommendations. All payment data were adjusted for inflation using the average consumer price index per year, standardised to 2016.10
Outcome measures
The primary outcome of interest was the type of error alleged with each paid malpractice claim. Paid claims were categorised into two main groups: diagnosis-related paid claims versus other paid claim types (eg, surgical, anaesthesia, medication, blood products, obstetrics, treatment, monitoring, equipment). Diagnosis-related paid claims were further categorised into subtypes including those resulting from failure to diagnose, delay in diagnoses, wrong diagnosis and other (all available within the NPDB).
Statistical analyses
Patient, provider and claim characteristics were analysed using descriptive statistics. Differences in patient, provider and claim characteristics between diagnosis-related claims and other paid claim types were assessed using Wilcoxon rank-sum test, t-tests and χ2 test, depending on the underlying distribution of the data. Multivariable logistic regression was used to assess the association between error type (diagnosis related vs other) and adverse events and identify patient and provider factors associated with diagnosis-related paid claims. Trends in incidence of diagnosis-related paid claims and median annual payment were assessed using Cochran-Armitage and non-parametric trend test. Results were expressed as percentages, medians and risk ratios, with corresponding 95% CIs; p<0.05 was considered statistically significant. Stata V.14 (StataCorp, College Station, TX) was used for analysis.
Ethical/regulatory considerations
The data used in this analysis were publicly available and were considered ‘not regulated’ by the institutional review board.
Result
A total of 109 903 paid malpractice claims filed between 1999 and 2011 were included in the analysis. Of these paid claims, 42.7% of associated allegations (n=46 937) occurred in the outpatient setting while 57.3% (n=62 966) occurred in inpatient, or both inpatient and outpatient settings. Regardless of location, diagnosis-related paid claims accounted for nearly one of every three paid claims (32.1%, n=35 231). Among diagnosis-related paid claims, 61.2% (n=21 549) occurred in the outpatient setting while 38.8% (n=13 682) occurred in the inpatient setting (figure 1). When examining inpatient paid claims, diagnosis-related paid claims were common (n=13 682, 21.7%) and second only to surgery-related paid claims (n=22 523, 35.8%). The proportion of all paid claims attributable to diagnostic errors decreased from 36.5% (n=2075) in 1999 to 29.1% (n=1591) in 2011 (p<0.001). Correspondingly, the proportion of inpatient diagnosis-related paid claims also decreased from 23.5% (n=779) for claims filed in 1999 to 21.2% (n=645) for those filed in 2011 (p=0.001) (figure 2). The most common subtypes of diagnosis-related paid claims were as follows: failure to diagnose (50%, n=6883), delay in diagnosis (27%, n=3750) and wrong diagnosis (5%, n=721). Notably, the proportion of inpatient diagnosis-related paid claims for failure to diagnose increased from 47.8% (n=369) in 1999 to 60.6% (n=389) in 2011 (p<0.001).
Population flow diagram. DO, doctor of osteopathic medicine; MD, doctor of medicine; NPDB, National Practitioner Data Bank
Proportion and number of diagnosis-related paid claims (by year of filing) in hospital settings
Outcomes associated with inpatient diagnosis-related paid claims
Of the 13 682 inpatient diagnostic error claims, 6475 (47.3%) were prompted by death and 4643 (33.9%) by disability. Although surgery-related paid claims were more prevalent overall (n=22 523), diagnosis-related paid claims were more often associated with mortality or disability than those related to surgery (81.3% (n=11 118) vs 45.9% (n=10 268), p<0.001) (table 1). Between 1999 and 2011, the proportion of diagnosis-related paid claims resulting in disability increased from 34.3% (n=265) to 38.2% (n=245) (p=0.05), whereas those resulting in death decreased from 49.6% (n=383) to 44.4% (n=285) (p=0.001). Changes in the proportion of diagnosis-related minor injury were not significant. During the same time period, the proportion of other types of paid claims resulting in minor injury increased from 31.2% (n=778) to 40.8% (n=974) (p<0.001), those resulting in disability decreased from 40.1% (n=999) to 28.1% (n=670) (p<0.001), and those resulting in death increased from 28.7% (n=716) to 31.1% (n=741) (p=0.04). Compared with those with other paid claim types, patients with diagnosis-related paid claims were 1.83 (95% CI 1.75 to 1.91; p<0.001) times more likely to be disabled and 2.33 (95% CI 2.23 to 2.43; p<0.001) times more likely to die than have a minor injury, after adjusting for patient and provider characteristics. Notably, a paid claim as a result of ‘failure to diagnose’ was associated with 1.16 times greater likelihood of death than a paid claim secondary to other subtypes of diagnostic error (95% CI 1.12 to 1.20, p<0.001).
Minor injury, disability and death related to malpractice claims (by allegation group), 1999–2011
Patient and provider factors associated with inpatient diagnosis-related paid claims
Baseline patient and provider characteristics associated with diagnosis-related paid claims are presented in table 2. The adjusted model, controlling for patient and provider characteristics as well as adverse event (minor, disability, death), is available in online supplementary table 1. Per the model, compared with those with other paid claim types, male patients were 1.17 times more likely to have a diagnosis-related paid claim than female patients (95% CI 1.13 to 1.20, p<0.001). Similarly, compared with other paid claim types, patients >50 years old were 1.04 times more likely to have a diagnosis-related paid claim than patients <50 years old (95% CI 1.01 to 1.07, p=0.010). Compared with other paid claim types, osteopathic physicians were 1.16 times more likely to have a diagnosis-related paid claim than allopathic physicians (95% CI 1.10 to 1.22, p<0.001). Compared with other paid claim types, physicians >50 years of age were 0.96 times more likely to have a diagnosis-related paid claim than those aged 20–49 (95% CI 0.93 to 0.99, p=0.010). Finally, compared with other types of paid claims, providers licensed in the northeast region were 1.11 times more likely to have a paid diagnosis-related claim than providers in other regions in the USA (95% CI 1.08 to 1.15, p<0.001). Notably, a total of 8.2% of physicians (n=1022) with one diagnosis-related paid claim also had a subsequent second diagnosis-related paid claim during the study period.
Supplementary file 1
Patient and provider characteristics associated with claims alleging diagnosis-related versus other types of errors
Economic impact associated with inpatient diagnosis-related paid claims
Diagnosis-related payments accounted for $5.7 billion (22.2%) of the $25.9 billion in inpatient payments between 2004 and 2016, and were second only to surgery-related payments. Importantly, 9% of physicians accounted for 50.9% ($2.9 billion) of payments. The median diagnosis-related payment was $250 682 (IQR $101 757–$527 873) and was the third highest among allegation groups, trailing obstetrics ($433 843.2; IQR $177 112–$946 729) and anaesthesia-related payments ($276 619; IQR $98 643–$634 515). Between 2004 and 2016, median diagnosis-related payments increased 31.2%, from $247 646 to $325 000 (p<0.001). In comparison, median payment for other error types increased 10.2%, from $222 246 to $245 000 (p=0.001) over the same period (figure 3). Regardless of claim type, median payments for claims resulting in minor injury, disability and death were $111 123 (IQR $41 929–$255 974), $441 657 (IQR $202 460–$888 904) and $225 599 (IQR $98 794–$490 854), respectively.
Trends in median payment for diagnosis-related versus other claims in hospital settings (by year, in US$)
Discussion
In this analysis of the NPDB, we found that inpatient diagnosis-related paid malpractice claims were common and associated with substantial morbidity and mortality. Compared with other paid claim types, diagnosis-related paid claims were most likely to be associated with death and (following surgery) cost. Furthermore, patient characteristics such as gender and age and certain provider characteristics such as age and training (osteopathic vs allopathic) were associated with greater risk of diagnosis-related claims than other paid claim types. Building on work performed by others,3 5 11 12 our paper provides new insights into outcomes and financial outcomes of inpatient diagnostic errors and paid malpractice claims. Future studies that focus on root causes of inpatient error ways to prevent them appear necessary.
Among diagnosis-related paid claims, failure to diagnose was the most common subtype and represented half of all diagnosis-related paid claims. Notably, the proportion of diagnosis-related paid claims attributable to failure to diagnose increased significantly over the study period. This finding is important because the failure to diagnose subtype was more likely than others to be associated with mortality. Several factors associated with the inpatient setting might explain why inpatient diagnosis is challenging. For instance, time constraints related to concurrent care of multiple patients, distractions associated with interruptions, paging or unpredictable workflow and diffusion of electronic medical records might impede the diagnostic process.1 13 Supporting this hypothesis, we found that trainees often implicated distractions or competing priorities such as patient care, educational conferences and duty hour restrictions as impairments to diagnosis in an ongoing study at two of our academic medical centres.14 Future studies that explore these aspects and identify ways in which to prevent or mitigate harm related to diagnostic errors would therefore be welcomed.
Interestingly, we found that certain provider characteristics were more likely to be associated with diagnosis-related paid claims. For example, compared with other paid claim types, providers over the age of 50 were 3.8% less likely to have a diagnosis-related paid claim than younger, presumably less experienced providers. Although of unclear clinical significance, this finding was unexpected, as previous studies have failed to demonstrate a correlation between inpatient provider experience and quality of care.15 16 Studies that engage more seasoned clinicians to understand their diagnostic process might provide unique insights in this regard. We also found that approximately 8% of physicians who had a diagnosis-related paid claim were the subject of a second diagnosis-related paid claim during the study period. Relatedly, in accordance with the Pareto principle,17 payments on behalf of approximately 9% of physicians were responsible for more than half of the over $5.7 billion in diagnosis-related payments. These findings have important policy implications, including aspects that regulate medical licensure and review of privileges to practise medicine for certain providers.
We observed that the proportion of inpatient diagnosis-related paid claims declined over the study period. Although it is impossible to determine the reason for this decline from our study, such decrease may reflect growing availability of technology—such as CT and MRI—that facilitate accurate and timely diagnoses. Alternatively, growing awareness and emphasis on diagnostic error as the next frontier for patient safety18 may explain this trend. From an economic perspective, while inflation-adjusted median payments for all claim types increased over the study period, diagnosis-related median payments rose at a rate disproportionate to other claim types. While median diagnosis-related payments did not substantially change between 2004 and 2014, they have shown year-over-year growth from 2014 to 2016. Higher median payments for paid claims resulting in disability in comparison to those resulting in minor injury or death may explain the disproportionate increase in diagnosis-related median payments as the proportion of diagnosis-related paid claims resulting in disability increased over the study period as compared with a decrease among other paid claim types. Ultimately, the decreasing proportion of diagnosis-related paid claims may do little to assuage physician fears of malpractice litigation in the setting of more expensive diagnosis-related payouts. This has important implications given the association between malpractice litigation and defensive medicine, possibly fuelling ‘overdiagnosis.’19–23
There are several limitations to our study. First, as suggested by others,24–26 the incidence of inpatient diagnostic error and its association with adverse events cannot be reliably estimated from malpractice database review, as not all cases of diagnostic error lead to ascertainable harm or a paid claim. Furthermore, non-payment of claims associated with error occurs much more frequently than payment of claims in which error did not occur.25 Thus, our study potentially underestimates rates and consequences of inpatient diagnostic errors and should be viewed as a conservative estimate. Investigation of diagnostic error using multiple modalities (medical record reviews, voluntary physician reporting, and so on) is needed to better understand both rates of diagnostic error and factors associated with greater patient harm and higher cost.27 28 Second, institutional payments are not included in the NPDB when allegations against individual providers are dropped as part of the agreement; therefore, these events are not reflected in the current analysis.29 Third, by using a malpractice claims database, it is possible we included more cases of error with more severe outcomes. Fourth, allegation categories (eg, diagnosis-related, surgery-related, treatment-related) are not defined in the database; thus, interpretation of outcomes by allegation type is difficult. Finally, the economic data included reflect those reported in the NPDB; whether additional payments or other forms of contribution occurred cannot be determined.
Limitations notwithstanding, our study also has important implications. First, the focus on inpatient diagnosis-related malpractice payments is important. Findings from this study clearly suggest that this is an area that warrants further attention given economic and clinical outcomes. Second, we identified multiple patient and provider factors more associated with diagnosis-related paid claims than other paid claim types. These factors might prove valuable when designing interventions, policy or payment strategies. Third, we found that failure to diagnose is common in hospital settings and appears to be increasing in incidence over time. This observation suggests the need for specific interventions when diagnoses are not clear or apparent, such as a team-based approaches or checklists for high-risk diagnoses.2 30
In conclusion, our study suggests the incidence and economic burden associated with diagnostic errors in the hospital is substantial. Further research focused specifically on diagnostic errors in hospitalised patients appears necessary if we are to improve this ‘next imperative for patient safety.’31
References
Footnotes
Contributors All authors contributed to the project conception, data analysis, manuscript drafting and revision. Each author has approved the final manuscript version and is accountable for all aspects of the work.
Competing interests None declared.
Patient consent Detail has been removed from this case description/these case descriptions to ensure anonymity. The editors and reviewers have seen the detailed information available and are satisfied that the information backs up the case the authors are making.
Provenance and peer review Not commissioned; externally peer reviewed.