OUP user menu

Adverse events in surgical patients in Australia

DOI: http://dx.doi.org/10.1093/intqhc/14.4.269 269-276 First published online: 1 August 2002


Objective. To determine the adverse event (AE) rate for surgical patients in Australia.

Design. A two-stage retrospective medical record review was conducted to determine the occurrence of AEs in hospital admissions. Medical records were screened for 18 criteria and positive records were reviewed by two medical officers using a structured questionnaire.

Setting. Admissions in 1992 to 28 randomly selected hospitals in Australia.

Study participants. Five hundred and twenty eligible admissions were randomly selected from in-patient databases in each hospital. A total of 14 179 medical records were reviewed, with 8747 medical and 5432 surgical admissions.

Main outcome measures. Measures included the rate of AEs in surgical and medical admissions, the proportion resulting in permanent disability and death, the proportion determined to be highly preventable, and the identification of risk factors associated with AEs.

Results. The AE rate for surgical admissions was 21.9%. Disability that was resolved within 12 months occurred in 83%, 13% had permanent disability, and 4% resulted in death. Reviewers found that 48% of AEs were highly preventable. The risk of an AE depended on the procedure and increased with age and length of stay.

Conclusion. The high AE rate for surgical procedures supports the need for monitoring and intervention strategies. The 18 screening criteria provide a tool to identify admissions with a greater risk of a surgical AE. Risk factors for an AE were age and procedure, and these should be assessed prior to surgery. Prophylactic interventions for infection and deep vein thrombosis could reduce the occurrence of AEs in hospitals.

  • adverse events
  • disability
  • injury
  • prevention
  • quality in health care
  • surgery

The prevalence of adverse events (AEs) in acute care hospitals is high and AE prevention is now a priority research area for improving the efficiency and effectiveness of health care [1]. Australia is one of three countries that have published national estimates of the magnitude and variety of AEs in acute care hospitals. The Harvard Medical Practice Study (HMPS) [2] gave an estimate of 3.7% for the incidence of AEs in New York using a random sample of 30 121 admissions in 1984; another US study, UTCO [3], made estimates for the incidence of AEs of 2.9% from 14 700 random admissions in 1992 for Colorado and Utah hospitals. The recently completed New Zealand study reported an AE rate of 12.9% after reviewing 6579 records in 13 hospitals [4] (see http://www.moh.govt.nz/moh.nsf). A pilot study in the UK obtained a rate of 6.7% from 480 medical records [5]. The Quality in Australian Health Care Study (QAHCS) [6] reviewed a random sample of 14 179 admissions to public and private hospitals in 1992, and reported that 16.6% of admissions are associated with an AE, and that 50.3% of AEs are associated with surgery.

The differences in reported rates, especially between the US studies with the low rates and the other two studies with higher rates, has generated discussion about the methodology of retrospective medical record review. A comparison of the results from UTCO and the QAHCS has recently been published [7,8]. Although the four studies used a two-stage process, with explicit criteria for the first stage and a similar definition for an AE for the medical review, the New Zealand and Australian studies were concerned with patient safety and quality of care, while the US studies had a greater interest in negligence. As a consequence of this difference in focus, the studies differ in three areas. The first stage review classified twice as many records as criteria-positive in QAHCS and New Zealand, which could be explained partially by the fact that the US studies limited the criteria for unplanned re-admissions to 6 months for ages >65 years. The HMPS and UTCO reported incidence rates, whereas ‘prevalence’, the proportion of admissions associated with an AE, was reported in QAHCS and New Zealand. The cutoff point for medical causation for the AE was set at two in QAHCS and New Zealand, but at four in the US studies. Adjusting for these factors still results in a comparison of UTCO and QAHCS of 3.2% and 10.6% [7]. Runciman et al. [8] suggest that this is due to differences in reporting AEs with minor disability. One of the areas reporting a difference is post-operative problems, i.e. wound infection and bleeding. The higher rate reported in QAHCS may make this study more relevant for the analyses in this paper.

The recent publication by the Institute of Medicine (IOM) on medical error [9] used the results from the HMPS and, by implication, the findings from the other studies, to highlight the magnitude of medical error. In particular, the IOM reported that 44 000–98 000 persons die each year in the US as a result of medical error. This has resulted in McDonald et al. [10] and Hayward and Hofer [11] questioning the validity of the results from HMPS and UTCO. A similar response was made to the publication of the QAHCS results, in which it was estimated that 18 000 deaths were associated with AEs in Australia in 1992. The key aspect when reporting the disability resulting from an AE in these retrospective record review studies is that no attempt has been made to determine what the outcomes may have been if no error had occurred. To do this would require estimating the probability of death given that the error was not made. This is a difficult exercise, and Hayward and Hofer [11] have attempted this on 111 randomly selected deaths in hospital. For retrospective record review studies, we prefer to report that the deaths were associated with an AE, rather than that the deaths were caused by the AE. Thus, 18 000 deaths would not be prevented if all errors were removed. This is supported by Brennan [12] in his discussion of post-operative haemorrhage, where transfer of a patient back to the operating room occurred even though there was no identifiable error made by the surgeon.

This distinction has relevance to this paper, where we report the disability resulting from AEs in surgical patients, including wound infections. A wound infection may be prevented by prophylactic antibiotics, but not always. Thus, a wound infection will be classified as a complication caused by medical care if prophylactic antibiotics were not used. However, the appropriate use of antibiotics (removing the ‘error’) does not necessarily mean that the infection would not have occurred. That is, we can report the number of wound infections that occurred, but we cannot infer that no infections would have occurred if the patients had received prophylactic antibiotics. In our study, disabilities resulting from surgical procedures that are associated with an AE suggest that some aspect of medical care was not provided, or in some cases was provided unnecessarily. From a quality in health care point of view, using record reviews to document the failure of key processes (appropriate provision of care) is more important than trying to estimate the potential gains in the outcomes, which can, in principle, only be done using randomized trials.

Leape [13] addresses the question of overestimation of error rates by recognizing that medical records often do not contain sufficient information to allow identification of all errors, and retrospective studies are likely to underestimate AEs/complications [14], and that prospective observational studies often report higher AE rates [15]. Leape also recognized that hindsight bias may influence the reviewer’s perspectives on deaths due to AEs, but goes on to explain that 86% were AEs that constituted a ‘major factor leading to the patient’s death’. He emphasizes that these errors are the result of ‘defects in the design and conditions of medical work’.

We report the magnitude of AEs for patients undergoing surgery, a class of patients including many who would be having an elective procedure rather than a non-elective admission for an acute episode of illness.

Materials and methods


The purpose of this study was to determine the AE rate for surgical patients, associated disability and preventability, and the major risk factors for an AE.


Definitions of AEs vary and the QAHCS used the following definition (which was also used in the three other studies mentioned above): an AE is ‘an unintended injury or complication which results in disability, death or prolongation of hospital stay, and is caused by health care management rather than the patient’s disease’ [6]. Details of the study methods have been reported previously [6]. A two-stage review process determined the occurrence of an AE associated with the index admission. The index admission was the admission sampled. In stage 1, the medical records were screened by trained registered nurses using 18 explicit criteria, and if one or more criteria was positive, the record was independently reviewed by two medical officers using a structured questionnaire that included questions on unintended injury or complication, disability, and causation. These three questions were used to determine the occurrence of an AE (stage 2). If there was disagreement between reviewers, the process was repeated jointly and presented to a third reviewer, and consensus obtained [6]. The reviewers were 21 medical officers who were specialists with a minimum of 10 years’ experience, and many were senior specialists. They included physicians (nine), anaesthetists (five), obstetricians (four), surgeons (two), and one paediatrician.


Twenty-eight hospitals were randomly selected from two states in Australia, in which admissions during 1992 were selected for review.

Study participants

Five hundred and twenty admissions were randomly selected by computer from in-patient databases in each hospital. Day-only admissions and patients in psychiatric wards were excluded. A total of 14 179 admissions were successfully reviewed, of which 8747 were medical and 5432 were surgical admissions.

Outcome measures

The proportion of admissions positive for each screening criterion was calculated.

The prevalence of AEs was defined as the proportion of admissions associated with an AE. Temporary disability was defined as resolved within 12 months, and permanent disability was defined as not resolved within 12 months. Preventability of an AE was defined as ‘an error in management due to failure to follow accepted practice at an individual or system level’; accepted practice was taken to be ‘the current level of expected performance for the average practitioner or system that manages the condition in question’ [6]. Three categories were used: not preventable, low preventability (some evidence for preventability but <50%), and high preventability (strong evidence of 50% or more).

AE rates were analysed by age and length of stay (LOS) to determine whether there was a significant association. χ2 analysis was used to test for the statistical significance of differences in AE rates. LOS (number of days for the index admission) and bed days attributable to the AE (the number of days of hospitalization due to the AE including readmission days) were collected by reviewers and used to indicate some of the costs associated with AEs.

The 14 179 admissions from the QAHCS database were classified according to their Australian National Diagnosis Related Group (DRG) as either surgical (5432) or medical (8747) (Figure 1) (DRG refers to the Australian version of DRGs) [16]. AE rates for the 20 most frequent surgical DRGs were calculated. Logistic regression was used to determine whether DRG, age, and LOS explained the observed variation in AE rates.

Figure 1

The data retrieval process. 1Operative is a category for an adverse event occurring in relation to an operation or within 30 days of an operation [6], and describes procedures performed in operating theatres, excluding endoscopic (coded as a medical admission), and day-only procedures.

System errors (those characterized by system problems, such as communication problems, delays, or equipment failures) and reasons for failure to prevent AEs were collected by reviewers. Categories of error were also calculated, including operative, system, diagnostic, therapeutic, drug, medical, fractures, and anaesthetic errors. Operative AEs were coded into surgical complications.

Performance errors were coded for operative AEs into categories previously used for studies of complex systems [1719]. The categories were: technical errors, skill-based errors, rule-based errors, knowledge-based errors, and violations. Eight ‘areas for attention’, which would help prevent the occurrence of operative AEs, were also coded and proportions within each category calculated. These categories were: quality assurance, education, system change, communication, credentialling, retraining, resources, and record keeping.


Forty three per cent of surgical admissions were positive for one or more of the 18 screening criteria and of these, 51% were found to have an AE (Figure 1). All 18 criteria were associated with a higher AE rate and the odds ratios ranged from 1.5 to 10.3 (Table 1). Five of these criteria are related to surgical procedures. These criteria were:

  1. unplanned return to operating theatre;

  2. hospital-acquired infection/sepsis;

  3. other patient complications (acute myocardial infarction, cerebrovascular accident, pulmonary embolus);

  4. unplanned removal/injury/repair of organ during surgery;

  5. development of neurological deficit.

View this table:
Table 1

Percentage of surgical admissions (n = 5432) positive for each criterion and the proportion of positive admissions with an adverse event

Eighteen per cent of surgical admissions were positive for one or more of these five criteria, and accounted for 49% of the AEs.

The prevalence of surgical admissions associated with an AE was 21.9% [95% confidence interval (CI) 20.8–23.0%], although an analysis of nine high-volume elective procedures, which represented 29.5% of all surgical admissions, had a slightly lower rate of 17.2%.

The degree of disability existing at the time of discharge was recorded for 1120 (94%) of the surgical AEs. AEs associated with minimal disability (resolved within 1 month) occurred in 46.9% of AEs, 36.1% of disabilities were resolved within 1–12 months, and 17.0% of disabilities were permanent, of which 4.0% were deaths.

Reviewers found that 47.6% of AEs were highly preventable, 31.4% had low preventability, and 20.8% were not preventable. The proportion of AEs with high preventability increased as the level of disability increased (from 45.5% for those with minimal disability to 64.4% who died; χ210 = 34.3, P < 0.0002).

Risk factors

The AE rates increased significantly with age (13.4% for ages 0–14 years and 28.9% for ages 65 years and over; χ24 = 83.1, P < 0.0001) (Table 2). However, this increase in age occurred primarily in AE rates with moderate disability and death (χ24 = 10.5, P = 0.03, and χ24 = 77.5, P < 0.0001, respectively). The proportion of surgical AEs with high preventability also increased with age (36% for ages 0–14 years, 54% for age 65 years and over; χ24 = 13.2, P = 0.01). There were no statistically significant differences between genders.

View this table:
Table 2

Number of surgical admissions and proportion associated with an adverse event, category of disability, and proportion with high preventability by age of patient

LOS was a statistically significant predictor for the occurrence of an AE (P < 0.0001). The AE rate for admissions with LOS of 1 or 2 days was 12%, but increased to 46% for the 838 admissions with LOS >12 days. Median LOS for admissions without an AE was 4 days, and for admissions with an AE was 8 days.

Bed days that were attributable to AEs increased with age from 3.8 days (age 0–14 years), 5.6 days (age 15–29 years), 5.8 days (age 30–44 years), 6.4 days (age 45–64 years), to 9.2 days (age 65+ years), with a mean of 7.1 days.

The AE rates varied from 4 to 38% for the 20 most common surgical DRGs (Table 3). DRGs with high AE rates were the more complex or major procedures.

View this table:
Table 3

The 20 most frequent surgical diagnosis-related groups and the proportion associated with an adverse event

When age (five categories), LOS and DRG (20 categories) were used to predict AE rates, only LOS and DRG were significant. Therefore, the procedure (DRG) and resulting LOS are the important risk factors for an AE.

DRG names may be listed as ‘with complications and co-morbidities’ (CCs), and none of these appeared among the 20 most common surgical DRGs. The combined AE rate for all 32 DRGs with CCs was very high (42.3%) compared with the rate for the 167 DRGs without CCs (20.7%). However, the CCs can either be present at admission or arise due to health care intervention, and hence they are an inadequate measure for determining the occurrence of an AE.

System errors were coded for 11.5% of all AEs, of which 76% were highly preventable. Over half (52%) of the system errors were coded as absence of or failure to use a policy, protocol, or plan, 17% were inadequate reporting, and 14% were associated with inadequate training or supervision of staff.

Reasons for the failure to prevent AEs were reported in 844 cases (73% of AEs), and explicit reasons were recorded in 452 cases (54%). The explicit reasons were as follows.

  1. Failure to take precautions to prevent accidental injury, 244 (29%). [Of these complications, the most common complications were wound infection (38), other wound problems (24), technical (20), and bleeding (13), and the remainder were widely distributed across categories.]

  2. Avoidable delay in treatment, 56 (7%).

  3. Failure to take an adequate history or physical examination, 45 (5%).

  4. Failure to employ indicated tests, 35 (4%).

  5. Failure to act upon the results of findings or tests, 37 (4%).

  6. The doctor or other health professional practicing outside his or her area of expertise (failure to consult/refer/seek assistance/transfer), 35 (4%).

Analysis of errors found that 870 (74.9%) of the AEs were a result of the operation and the others included system errors, diagnostic errors, therapeutic errors, drug errors, medical errors, fractures, and anaesthetic errors. In the operative AEs, the following complications were coded: wound infection (2.1%), bleeding (1.4%), wound problems (1.2%), deep vein thrombosis/pulmonary embolism (0.3%), and pneumonia (0.2%). Performance errors for operative AEs are listed in Table 4. Eight areas for attention to prevent recurrence of AEs and their association with permanent disability, death, and preventability are shown in Table 5.

View this table:
Table 4

Performance errors that contributed to operative adverse events

View this table:
Table 5

Areas for attention to prevent recurrence of operative adverse events by proportion with permanent disability, deaths, and high preventability


This study is one of several that have specifically addressed surgical AEs [14,2023]. Gawande et al. [20] reported that 66% of all AEs were surgical, compared with 51% in this study. Reported AE rates vary from 0.6% [22] in the field of general surgery, to 6.9–33% for different procedures [21] and 17% for patients having one or more complications [23]. These studies use different methods and hence the variation is expected.

A comparison of surgical AEs (1190) with medical AEs (1163) shows that surgical admissions had a higher AE rate (21.9%) than medical admissions (13.3%). This is despite the fact that medical patients are primarily non-elective admissions.

Of the disabilities, 83% were temporary, being resolved within 12 months. We found that 47.6% of all surgical AEs were highly preventable. Other studies claim higher rates of preventability ranging from 52% [24], to 54% [20] and 74% [25].

This study confirms that age [20,21] is a significant risk factor for an AE, and should influence the decision on whether to operate. The higher AE rates reported for some DRGs indicate a higher risk associated with these procedures. They also provide impetus for the adoption of preventative measures. However, for the 20 most common DRGs, LOS is an important predictor, whereas age is not. The finding that surgical AEs were associated with increased LOS is supported by other studies [21,2629]. In our study it cannot be determined whether this increased LOS in the index admission is a result of an AE, or whether it is due to the co-morbidity of the patient or to the complexity of the procedure. However, the 7.1 bed days attributable to an AE indicate that some of the difference is due to the AE and provides an estimate of the increased costs associated with AEs.

Errors in clinical management are often a feature of the complex systems that characterize the clinical environment within a hospital [13,19,30,31]. Analysis of the errors that contributed to surgical AEs in our study identified that 11% of AEs were associated with system error (not the fault of an individual) and could not be readily attributed to one source or another. This contrasts with the 85% found in manufacturing [3234], and may reflect a lack of documentation of system errors in the medical record. For example, information such as busy shifts/overwhelming caseloads or admissions, shortages of staff, delayed test results, and untimely receipt of or inaccuracy in communication is not yet routinely recorded in patients’ medical records. Reasons for failure to prevent AEs were only coded explicitly in 54% of cases. Where systems can be improved to ensure safe routines and precautions are followed, many of these can potentially be prevented.

The high rate of complications such as infection, bleeding, and deep vein thromboses/pulmonary emboli [20] raise the issue of whether evidence-based prophylactic protocols and guidelines have been successfully implemented.

Performance errors could be prevented by establishing safe systems in the hospital. Slips and lapses may be minimized by improved rostering to reduce fatigue and excessive overtime. Education and planning for appropriate supervision and referral has the potential to reduce technical errors. Improvements in communication timeliness and accuracy provide a basis for better decision-making and better compliance with desired protocols. The areas for attention identified support these concepts of continued monitoring and review, system changes, education and retraining, communication, and others.

We recognize the following limitations of the study. It is likely that the rate reported is an underestimate, as discussed previously, due to retrospective record review and the lack of some information in medical records [13]. Inter-rater agreement among reviewers about severity of an AE and preventability is recognized as modest with a kappa score of 0.55, which is evidence of the difficulty of rendering judgements about causality [15]. The study design did not include specific collection of co-morbidity to generate a risk profile for AEs. The data are not recent, and changes in operative techniques and in post-operative management have occurred during this time; however, we believe that these are unlikely to substantially affect the results. The high AE rate for surgical procedures supports the need for monitoring and intervention strategies. The reporting of some of the 18 screening criteria provides a screening tool to identify admissions with a greater risk of a surgical AE. Risk factors for an AE, including age and procedure, should be assessed before surgery. Prophylactic interventions for infections, deep vein thromboses/pulmonary emboli, and bleeding, which are based on therapeutic guidelines, could reduce the occurrence of surgical AEs in hospitals.

Actions taken to reduce the occurrence of AEs will reinforce the importance of ‘primum non nocere’ [35] and will form an important part of the implementation of the clinical governance concept [36,37]. Government initiatives in the US and Australia directed at reducing AEs should focus on priority areas such as surgical admissions and prevention of post-operative infection and other preventable AEs.


This study was conducted using the QAHCS database, compiled by the following researchers in 1995: Ross McL. Wilson, William B. Runciman, Robert W. Gibberd, Bernadette T. Harrison, Liza Newby, and John D. Hamilton. The study was supported by a grant provided by the Ministerial Advisory Committee on Quality in Health Care, NSW Department of Health, Australia. This grant also provided a scholarship for A.K.K.


  • Address reprint requests to A. Spigelman, Discipline of Surgical Science, University of Newcastle, c/o John Hunter Hospital, Locked Bag 1, Hunter Region Mail Centre, Newcastle, NSW 2310, Australia. E-mail: allan.spigelman{at}newcastle.edu.au


View Abstract