Abstract
Objective:
The primary objective of this study was to quantify the differences in the prevalence rate and costs of hospital medical errors between the general population and an elderly population aged ≥65 years.
Methods:
Methods from an actuarial study of medical errors were modified to identify medical errors in the Premier Hospital Database using data from 2009. Visits with more than four medical errors were removed from the population to avoid over-estimation of cost. Prevalence rates were calculated based on the total number of inpatient visits.
Results:
There were 3,466,596 total inpatient visits in 2009. Of these, 1,230,836 (36%) occurred in people aged ≥ 65. The prevalence rate was 49 medical errors per 1000 inpatient visits in the general cohort and 79 medical errors per 1000 inpatient visits for the elderly cohort. The top 10 medical errors accounted for more than 80% of the total in the general cohort and the 65+ cohort. The most costly medical error for the general population was postoperative infection ($569,287,000). Pressure ulcers were most costly ($347,166,257) in the elderly population.
Limitations:
This study was conducted with a hospital administrative database, and assumptions were necessary to identify medical errors in the database. Further, there was no method to identify errors of omission or misdiagnoses within the database.
Conclusions:
This study indicates that prevalence of hospital medical errors for the elderly is greater than the general population and the associated cost of medical errors in the elderly population is quite substantial. Hospitals which further focus their attention on medical errors in the elderly population may see a significant reduction in costs due to medical errors as a disproportionate percentage of medical errors occur in this age group.
Introduction
The prevalence of medical errors places additional financial burden on hospitals in the US in a very direct way. In 2010, the Affordable Care Act established an array of financial incentives for quality careCitation1 and financial disincentives for preventable medical errors. This legislation followed on the heels of action taken by the Centers for Medicare & Medicaid Services to eliminate reimbursement for certain hospital-acquired infections.
Hospitals have been impacted by medical errors for decadesCitation2,Citation3, but it was To Err Is Human: Building a Safer Health SystemCitation4, the 2000 seminal report from the Institute of Medicine, that led hospitals to become more pro-active in addressing this issue. Despite the growing attention paid to medical errors, much of the discussion in the literature focuses on the prevalence and high cost of adverse eventsCitation2,Citation5–9 and unreimbursed hospital-acquired infectionsCitation10,Citation11, rather than medical errors specifically.
To expand research on the subject of medical errors, the term must first be defined. According to the Institute of Medicine report, a medical error is ‘the failure of a planned action to be completed as intended (error of execution) or the use of a wrong plan (error of planning) to achieve an aim’ (p. 4)Citation4. In other words, a medical error is preventable. It is differentiated from an adverse event, which refers to ‘an injury caused by medical management rather than the underlying condition of the patient. Adverse events may or may not be preventable, but a sizeable portion of them result from medical error’ (p. 28) Citation4. A similar term, “medical injuries”, may be defined as “any adverse events which occur due to medical intervention” (p. 5)Citation12.
Terminology is important because it is the underpinning of The Economic Measurement of Medical Errors, a pivotal study published in 2010 sponsored by the Society of Actuaries’ Health Section and conducted by Milliman, Inc.Citation12,Citation13. This study, which describes medical error as a sub-set of medical injury, used MarketScan®, a large health insurance claims database, as well as the Medicare Supplemental and Coordination of Benefits claims database, to determine the economic burden of medical errors on society. However, this study did not look at hospitals' costs related to medical errors. The literature on the prevalence and cost of medical errors in the hospitalized Medicare population is minimal. A handful of studies sponsored by the US Department of Health and Human Services (HHS) looked at prevalence and cost of adverse events in Medicare beneficiariesCitation14,Citation15, and there are annual studies on patient safety in the Medicare population from Health Grades, Inc. (Denver, CO), an independent healthcare ratings organizationCitation16.
Because the Milliman study noted that nearly half the expenditures of medical errors were linked to the age 65 and older demographic, and the literature identified the steep cost of adverse events in this group, gaining a more detailed understanding of medical errors in this population can offer insight into anticipated hospital costs that may not be reimbursed by Medicare. Also, it would be useful for hospitals to have data showing quantifiable differences in prevalence and type of medical error between the elderly and general populations, as this information will have cost implications and may signal the areas most in need of improvement.
The primary objective of this study was to determine the magnitude of the differences in the prevalence rate of medical errors between the general population and a population aged 65 and older. A secondary objective was to determine the magnitude of the differences in median costs of medical errors in these populations.
Methods
This study adapted the methodology developed by Milliman to identify medical injuries and estimate medical errors using the Premier Hospital Database. Once the medical errors were identified, the prevalence rates and estimated annual costs of the top 10 medical errors were calculated. Explanation of the data source, subjects, and further description of the specific methods follow below. The analyses were performed using SAS® software, version 9.2 (Cary, NC). Univariate t-tests and multivariable statistical techniques, including propensity matching, were utilized.
Subjects and database
For a hospital visit to qualify as an injury visit, at least one of the 97 injury groupings must have occurred at that specific visit (Appendix A). Visits with four or more unique injuries occurring at that visit were removed from analysis, and visits with more than one injury occurrence were only counted once in the general cost table to avoid over-estimation of cost. However, each injury was counted for the prevalence calculations. Hospital visits were identified from the clinical and billing data from the Premier hospital databaseCitation17,Citation18. The time period of October 1, 2008 through March 31, 2010 was examined to identify injury visits and costs for the year 2009. The 2008 and 2010 data was used to understand if patients had prior or post visits related to their medical error.
The Premier hospital database contains clinical and utilization information from over 600 US facilities (∼10% of US hospitals) and includes more than 45 million inpatient discharges and more than 210 million hospital outpatient visits from acute care facilities, ambulatory surgery centers, and clinics across the US.
Identifying injuries and errors
The Milliman study identified medical injuries by International Classification of Diseases, Ninth Revision (ICD-9) codes (Appendix A), and estimated how often each type of injury was likely to be associated with a medical error rather than a consequence of the underlying disease. These medical injuries were categorized into 97 injury groupings. Each grouping was classified based on the likelihood that they were associated with a medical error. The five classifications were: 0–10%; 10–35%; 35–65%; 65–90%; and >90% (Appendix B). The mid-point of each range was applied to the frequency of each medical injury to establish the count of medical errorsCitation12.
The current study applied the ranges and mid-points of medical errors established in the Milliman study to the data from the Premier database. All 97 injury groupings were re-verified by an outside medical coding group to ensure that no major changes had occurred since the Milliman list was developed.
After identifying visits for each type of injury, the likelihood that the specific injury was caused by a medical error was estimated. The final frequency of a specific type of medical error was estimated by multiplying the calculated frequency of the specified type of injury by the mid-point of the error percentage category. For example, the occurrence of a pressure ulcer was determined at a greater than 90% probability of medical error. Using the mid-point of this range, we estimated that 95 out of 100 pressure ulcers were medical errors.
Prevalence calculation
The general and elderly population prevalence rates were calculated by dividing the estimated number of medical errors and the total number of inpatient hospital visits for the respective groups. Chi-square tests were performed to assess statistically significant differences in the prevalence rate. The top 10 medical errors for both the general and elderly groups were reported.
Cost analysis
For each injury visit, a matched non-injury control was chosen to compare the difference between direct medical costs to hospitals. Non-injury control groups were established by propensity score matching which reduces the bias of confounding variables not associated with the injury. The propensity score for each subject was estimated based on gender, age group, severity of illness, risk of mortality, admission type, major comorbidities, and hospital characteristics. After propensity matching, the differences between the injury and non-injury populations were removed. Visits incurring costs below $300 or above $300,000 were removed from the cost analysis as they were determined to be outside the normal range of inpatient costs. The cost per error was estimated as the difference in cost between the injury cohort and the non-injury control cohort. T-tests for statistically significant differences in direct cost were performed.
National cost projections
This study utilized a projection methodology developed by Premier and validated by the US Food and Drug Administration to estimate the US aggregate costs of medical errors in the general and elderly populations. Hospital-specific projection weights have been created using geographic region, bed size, teaching status, and urban/rural status. Based on these weights, the prevalence of inpatient discharges in the Premier database were projected to create a nationally representative number of inpatient discharges for the denominator. The Premier weights were then applied at the medical injury level and the projected counts of medical errors were estimated as described previously. Annual costs associated with an error from the propensity matched population were applied to the projected injury and error rates to estimate the cost of errors nationally.
Results
In the Premier database, there were 3,466,596 total inpatient visits in 2009. Of these, 1,230,836 (36%) occurred in people aged 65 and older. Of the injuries and errors in the general cohort, the elderly cohort accounted for 96,756 (57%) unique medical errors and 330,247 (50%) unique injuries. This amounted to a prevalence rate of 49 medical errors per 1000 inpatient visits in the general cohort and a prevalence rate of 79 medical errors per 1000 inpatient visits for the elderly cohort ().
Table 1. Prevalence rate of medical errors, 2009.
Prior to propensity score matching, the patient and hospital characteristics of the general injury population were heterogeneous compared to the non-injury population (Appendix C). The general injury population was much more likely to be categorized as a major or extreme severity of illness (57.4% vs 25.1%) and be at a high risk (score of 3 or 4) of mortality (41% vs 14.9%). Consistent with the severity of illness and risk for mortality, the general injury population had a substantially higher percentage of patient visits who presented at the hospital with comorbidities. In particular, the percentage of heart-related comorbidities were much higher than in the non-injury population.
Similar to the general population, the elderly injury and non-injury population was heterogeneous prior to the propensity matching (Appendix C). Over 66% of the elderly injury population had a severity of illness measure of major or extreme compared to 43% of the non-injury elderly population. Fifty-five percent of the elderly injury population was considered at high risk of mortality (score of 3 or 4) compared to 32% of the elderly non-injury population. However, there was little difference in the comorbidities between the two groups. Hypertension was higher in the elderly non-injury group compared to the elderly injury group (57% vs 51%), as was cardiac dysrhythmias (37% vs 30%).
After propensity matching, the top 10 medical errors accounted for more than 80% of the total number of medical errors in the general and the aged 65 and older cohorts. The top three medical errors—pressure ulcer, post-operative infection, and iatrogenic hypotension (inadvertent low blood pressure)—were equal in both the general and elderly cohorts. The fourth through ninth medical errors were the same in the two cohorts, but occurred in different orders. The tenth most common medical error was abnormal reaction due to surgery without mentioning of misadventure in the general cohort; but pneumothorax was number 10 in the elderly group. With the exception of infection due to central venous catheter, the prevalence rate of medical errors was substantially higher in the elderly group for the remaining medical errors ( and ).
Table 2. Top 10 general population medical errors ranked by prevalence, 2009.
Table 3. Top 10 elderly population medical errors ranked by prevalence, 2009.
The top 10 medical errors for both cohorts were extrapolated to the US population and their associated median costs were estimated ( and ). The median costs represent 2009 annual hospital incurred costs. The most costly medical error for the general population was post-operative infection ($569,286,000), but pressure ulcers were most costly ($347,166,000) in the elderly population. The median costs of injuries were similar between the two populations, with the exception of accidental puncture or laceration during a procedure, which was 38% higher in the elderly population ($2717) than the general population ($1974).
Discussion
This analysis found that the prevalence of medical errors occurring in the elderly population was substantially higher than the general population (79 medical errors per 1000 inpatient visits compared to 49 medical errors per 1000 inpatient visits). After propensity matching, the cost of the medical errors was found to be quite substantial. The most costly medical error in the general population was post-operative infections ($569 million) and the most costly medical error in the elderly was pressure ulcers ($347 million). The US Census forecasts the size of the elderly population as increasing from 13% today to 19% by 2030Citation19. As the US population continues to age, the associated costs of these medical errors are likely to rise.
For comparison, the results of the Milliman study showed that, in 2008, 7% of inpatient admissions in the US led to some type of medical injuryCitation12. The mean societal cost per error was calculated at ∼$13,000, resulting in a total annual cost to the US economy of $19.5 billion. Direct costs of medical errors were estimated at $17.1 billion.
Other studies that examined medical errors found qualitatively similar results to this study. A 2010, North Carolina study determined the rate of error to be 30 per 1000 inpatient admissionsCitation7. Another study examining adult hospitalizations in New York determined that 37 per 1000 of all hospitalizations were associated with adverse events and negligenceCitation5. A 2005 study of patients seen by the Veterans Health Administration found 20 per 1000 of hospitalization resulted in a patient safety event as defined by the Agency for Healthcare Research Quality’s Patient Safety IndicatorsCitation20. These studies have different study designs and populations, making a direct comparison difficult, but they provide context that the problem of medical errors is a significant burden on hospitals.
Medicare has recently restricted reimbursement for the treatment of ‘never events’, which are defined as events that should never occur in a hospital. Four of the top 10 medical errors in the general and elderly population can be linked to the list of never eventsCitation21. Pressure ulcers and catheter-associated urinary tract infections are directly listed. Accidental puncture or laceration and substances causing adverse effects in therapeutic use can be linked to Medicare never events depending on the specific circumstances. This study has found that pressure ulcers and catheter-associated urinary tract infections accounted for over $570 million in hospital costs in 2009, of which over $400 million occurred in the elderly population. The recent focus on patient outcomes through regulatory efforts and the legislative efforts of the Affordable Care Act of 2010 have demonstrated that hospitals must continue to analyze the causes of medical errors and seek new ways to reduce their occurrence.
This study suggests that substantial progress in reducing the hospital’s cost of medical errors can be made through increased attention to programs that specifically target an elderly population. While the elderly accounted for only 36% of inpatient hospitalizations, more than half of the injuries and errors occurred in this population. Although reasons for the higher rate of errors among the elderly cannot be ascertained from this database, one contributing factor may be the longer mean length of stay in this cohort (7.7 days compared to 7.0 days; data not shown). However, the longer length of stay may have been due to the underlying cause of the visit rather than resulting from the injury. An additional contributing factor may include the increased likelihood of comorbid conditions among the elderly. Finally, younger individuals may not present adverse outcomes of a medical error, thus medical errors may go unreported at higher rate compared to the elderly.
Study limitations
This study had several limitations. First, the data used to identify medical errors came from a hospital system administrative database, rather than a database expressly designed to track medical injuries and errors. Therefore, the hospital claims data may lack the clinical detail necessary to definitively define a medical error. This study relied on hospital coding data, which may introduce measurement error due to different coding practices of hospitals. This analysis only contained data from inpatient hospital facilities; therefore, injuries and costs that occurred in other levels of care were not captured. Furthermore, the direct cost of medical errors to hospitals in this study was likely a conservative estimate. Visits that contained four or more unique injuries per visit or had costs exceeding $300,000 per visit were excluded from the analysis. The cost of injuries may also have been under-estimated, as only unique injuries per visit were included in the analysis. Other costs, such as lost work time, were not considered. Finally, the method employed in this study did not account for a misdiagnosis or negligence on behalf of the hospital (errors of omission).
Conclusions
This study provides important insights into the epidemiology of medical errors in the general and, in particular, elderly populations hospitalized in a US inpatient setting. The impact of medical errors on hospitals is large and continues to rise. The associated cost of medical errors in the elderly population is quite substantial. Hospitals which further focus their attention on medical errors in the elderly population may see a significant reduction in costs due to medical errors, as a disproportionate percentage of medical errors occur in this age group.
Transparency
Declaration of funding
This study was funded by GE Healthcare. The publication of study results was not contingent on the sponsor’s approval or censorship of the manuscript.
Declaration of financial/other relationships
Peter Mallow is an employee of S2 Statistical Solutions, Inc., which is the paid consultant to GE Healthcare. Bhavik Pandya and Ruslan Horblyuk were employees of GE Healthcare, the study sponsor at the time the study was conducted. Harold Kaplan has been sponsored by and is a consultant/advisor to GE Healthcare. JME Peer Reviewers on this manuscript have no relevant financial relationships to disclose.
Acknowledgments
No assistance in the preparation of this article is to be declared.
Supplementary Material
Download PDF (356.2 KB)References
- Kocher R, Emanuel EJ, DeParle NA. The Affordable Care Act and the future of clinical medicine: the opportunities and challenges. Ann Intern Med 2010;153:536-9
- Gawande AA, Thomas EJ, Zinner MJ, et al. The incidence and nature of surgical adverse events in Colorado and Utah in 1992. Surgery 1999;126:66-75
- Mills DH. Medical insurance feasibility study. A technical summary. West J Med 1978;128:360-5
- Kohn LT, Corrigan J, Donaldson MS. To err is human: building a safer health system. Washington: National Academy Press, 2000
- Brennan TA, Leape LL, Laird NM, et al. Incidence of adverse events and negligence in hospitalized patients. Results of the Harvard Medical Practice Study I. N Engl J Med 1991;324:370-6
- de Vries EN, Ramrattan MA, Smorenburg SM, et al. The incidence and nature of in-hospital adverse events: a systematic review. Qual Saf Health Care 2008;17:216-23
- Landrigan CP, Parry GJ, Bones CB, et al. Temporal trends in rates of patient harm resulting from medical care. N Engl J Med 2010;363:2124-34
- Naessens JM, Campbell CR, Huddleston JM, et al. A comparison of hospital adverse events identified by three widely used detection methods. Int J Qual Health Care 2009;21:301-17
- Paradis AR, Stewart VT, Bayley KB, et al. Excess cost and length of stay associated with voluntary patient safety event reports in hospitals. Am J Med Qual 2009;24:53-60
- Fuller RL, McCullough EC, Bao MZ, et al. Estimating the costs of potentially preventable hospital acquired complications. Health Care Financ Rev 2009;30:17-32
- Milstein A. Ending extra payment for “never events"–stronger incentives for patients' safety. N Engl J Med 2009;360:2388-90
- Shreve J, Van Den Bos J, Gray T, et al. The economic measurement of medical errors: sponsored by Society of Actuaries’ Health Section. Schaumberg, IL: Milliman, 2010
- Van Den Bos J, Rustagi K, Gray T, et al. The $17.1 billion problem: the annual cost of measurable medical errors. Health Aff (Millwood) 2011;30:596-603
- Levinson DR. Adverse events in hospitals. Case study of incidence among Medicare beneficiaries in two selected counties. Washington, DC: Department of Health and Human Services. Office of Inspector General, 2008; https://www.oig.hhs.gov/oei/reports/oei-06-08-00220.pdf. Accessed August 16, 2012
- Levinson DR. Adverse events in hospitals: national incidence among Medicare beneficiaries. Washington, DC: Department of Health and Human Services. Office of Inspector General, 2010; https://oig.hhs.gov/oei/reports/oei-06-09-00090.pdf. Accessed August 16, 2012
- Reed K, May R. HealthGrades patient safety in American hospitals study. Denver, CO: Health Grades, Inc., 2011; https://www.cpmhealthgrades.com/CPM/assets/File/HealthGradespatientSafetyInAmericanHospitalsStudy2011.pdf. Accessed August 16, 2012
- Premier. Transforming Healthcare, https://www.premierinc.com/wps/portal/premierinc/public/transforminghealthcare/improvingperformance/servicesprograms/researchservices/!ut/p/b0/04_SjzQyMDAxtTAwMrTUj9CPykssy0xPLMnMz0vMAfGjzOJNzQzMnJwMHQ3cvQwsDDxDzP1MnX18Df29jPVzoxwVAc9243w!/?1dmy&urile=wcm%3apath%3a/general+public/events/6eb780fb-ff2d-4c3b-9085-cfc51fc1ede6. Accessed 9 Oct 2013
- Premier Quality, https://www.premierinc.com/wps/portal/premierinc/public/transforminghealthcare/improvingperformance/quality. Accessed 9 Oct 2013
- Vincent GK, Velkoff VA. The Next Four Decades: The Older Population in the United States: 2010 to 2050. Washington, DC: US Census Bureau, 2010; http://www.census.gov/prod/2010pubs/p25-1138.pdf. Accessed August 16, 2012
- Rosen AK, Rivard P, Zhao S, et al. Evaluating the patient safety indicators: how well do they perform on Veterans Health Administration data? Med Care 2005;43:873-84
- National Quality Forum. List of SREs. Washington, DC: National Quality Forum, 2011; http://www.qualityforum.org/Topics/SREs/List_of_SREs.aspx#sre1. Accessed August 29, 2013
Appendix A. ICD-9 codes for injuries associated with an error
Table A1. ICD-9 codes for injuries associated with an error.
Appendix B. Percentage of occurrences with error
Table A2. Percentage of occurrences with error.
Appendix C. Injury and no injury population prior to propensity matching
Table A3. Injury and no injury population prior to propensity matching.