This is the Accepted Manuscript version. This version is defined in the NISO recommended practice RP

Version This is the Accepted Manuscript version. This version is defined in the NISO recommended practice RP-8-2008 http://www.niso.org/publications/rp/ Suggested Reference Brown, P. M., Mcarthur, C., Newby, L., Lay-Yee, R., Davis, P., & Briant, R. H. (2002). Cost of Medical Injury in New Zealand: A retrospective cohort study. Journal of Health Services Research and Policy, 7(Suppl. 1), S29-S34. doi:10.1258/135581902320176449 Copyright Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated. Previously published items are made available in accordance with the copyright policy of the publisher. https://au.sagepub.com/en-gb/oce/the-green-route-%e2%80%93-open-accessarchiving-policy http://www.sherpa.ac.uk/romeo/issn/1355-8196/ https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm

Cost of Medical Injury in New Zealand: Results from the NZQHS Abstract: Objectives: Identify the cost of treating medical injury associated with hospital admissions in New Zealand and the patient characteristics of costly adverse events. Methods: As part of the New Zealand Quality in Healthcare Study (NZQHS), a retrospective examination of medical records in 13 public hospitals identified the additional procedures (e.g., surgery) and hospital bed-days attributable to the adverse events. The prices charged to international patients were used to estimate the cost of the healthcare resources. Results: The 850 adverse events identified in the NZQHS cost an average of $NZ10,264 per patient. For New Zealand, adverse events are estimated to cost the medical system over $NZ870 million, of which over $NZ590 million went toward treating preventable adverse events. The results suggest that up to $.30 of every dollar spent in a public hospital goes toward treating an adverse event. The results also suggest that older patients, neonates and those with moderately serious co-morbidity tended to have more costly adverse events. Conclusions: Adverse events are a significant drain of healthcare resources in New Zealand. These findings suggest that substantial resources could be saved from eliminating preventable adverse events. Submitted January 2002 Revised March 2002 Paul M Brown, PhD University of Auckland Colin McArthur, MBCHB Auckland District Health Board Lynette Newby, RN Auckland District Health Board Roy Lay-Yee, MA University of Auckland Peter Davis, PhD Christchurch School of Medicine Robin Briant, MBCHB, MD, FRACP University of Auckland Correspondence to: Paul M Brown, Senior Health Economist, Centre for Health Services Research and Policy & Department of Community Health, University of Auckland, Private Bag 92019, Auckland, New Zealand, pm.brown@auckland.ac.nz

1. Introduction: A number of studies have demonstrated that medical injuries (also called adverse events) represent a significant drain on health care resources. Johnson et al (1992) found that adverse events cost the New York health care system $161 million (in 1989 dollars); Wilson et al (1995) estimated that adverse events in Australia cost the government over $900 million in 1995; Thomas et al (1999) concluded that adverse events cost the medical system in the states of Utah and Colorado over $348 million; and Vincent et al (2001) found that errors in Britain cost over 2400 per adverse event (see Table 1). The current study examines the healthcare cost of adverse events in New Zealand. The recently completed New Zealand Quality in Healthcare Study (NZQHS, Davis et al, 2001) identified the occurrence, impact and preventability of adverse events associated with hospital admissions in 1998. Interest in identifying the cost of adverse events to the New Zealand healthcare system stems from the recognition that the money spent on adverse events is diverted away from other health care services. As in previous studies, the study highlights the cost of preventable adverse events as this provides an indication of the savings that might be available from successful interventions. The current study also examines the patient characteristics associated with costly adverse events. To date, no study has modelled the relationship between the patient characteristics and the cost of the adverse event (although Rigby et al, 1999, examined the relationship between cost and different types of adverse events). We estimate a predictive model of the healthcare costs associated with adverse events using information available to medical professionals about the patient (i.e., gender, age, ethnicity, social economic status and degree of co-morbidity) and the severity of the adverse event. This information can assist hospital staff and other health professionals in identifying patients who are likely to require lengthy hospitalisations or extensive procedures. Such information is particularly useful when developing interventions aimed at reducing adverse events. 2. Method The NZQHS sought to determine the occurrence, impact and preventability of adverse events in New Zealand s healthcare system. A retrospective, two-stage review 1

was carried out on 6,579 medical records in 13 public hospitals with 100 or more beds in New Zealand (see Davis, et al, 2001, for more detail). The sample was a representative selection of approximately 1% (1 in 100) of all inpatient admissions (excluding day, psychiatric and rehabilitation-only cases) during 1998. To determine whether an admission was associated with an adverse event, a team of trained registered nurses and medical officers applied the standardised protocol used in the Australian (Wilson et al, 1995) and New York (Brennan et al, 1991, Leape et al, 1991, Localio et al, 1991) studies. An adverse event was operationally defined as an unintended injury resulting in temporary or permanent disability, including increased length of stay and/or financial loss to the patient, that was caused by healthcare management rather than the underlying disease process. The event was deemed preventable if the medical officers agreed that the error in health care management was due to a failure to follow accepted practice at the individual or system level. The adverse event could have originated as the result of care inside the hospital or outside (e.g., general practice), but all were associated with a sampled hospital admission in 1998. Twenty percent of the adverse events occurred prior to 1998 but were part of the sampled admissions of 1998. Because the cost of these adverse events was associated with the admission of 1998, they were included in the sample. The general approach to estimating the cost of adverse events is to identify the additional healthcare resources used as a result of the adverse event and then apply a charge or cost to those resources. In the current study, the procedure for identifying and estimating the additional healthcare resources was as follows: Number of additional bed days attributable to the adverse event Following the initial determination of the adverse event, the medical officers estimated from the medical records the additional number of bed days attributable to the event. Identification of hospital ward Using the medical officer s description of the adverse event, an investigator (CM) and a research nurse identified the ward or area in the hospital where the patient spent the extra bed days (e.g., General Medical, Intensive Care Unit, Oncology). When patients spent time in multiple wards, the specific number of days in each ward was recorded. Additional procedures Using the medical officer s description, an investigator (CM) and research nurse identified the additional procedures that were undertaken 2

as a result of the adverse event. Only those procedures deemed not normally performed as part of routine stay on the specific ward were included (e.g., additional surgery or specialised procedures). The medical officers applied the criteria used in the Australian and New York studies for identifying the number of bed days attributable to the adverse event (see Davis, et al, 2001, for evidence on the validity and reliability of the medical officer s assessments). As a check of the consistency of the assessments of the ward and additional procedures, the investigator (CM) and research nurse compared assessments of 10% of the cases. There were no significant areas of disagreement. Cases deemed by the investigator and research nurse to be complicated were discussed jointly throughout the investigation. The procedure produced a list of the number of the bed days by hospital ward and additional procedures deemed not part of the normal care on the ward. The next step was to assign a cost or charge to these resources. Due to the retrospective nature of the data, it was not possible to identify the specific consumable resources attributable to each patient (see Jackson, 2000, and Graves and Brown, 2001, for discussion of approaches to estimating hospital costs). Instead, the current study used the price charged by Auckland Hospital to international patients for a bed day on the specified ward and for the specified procedures in 2001 as an estimate of the value of the resources associated with the adverse event. All prices are reported in New Zealand dollars. The total healthcare expenditure associated with an adverse event was calculated for each patient by summing the total cost of his or her procedures and the total bed day costs in each ward. The procedure yielded a cost of the adverse event for each patient. To estimate the total cost to the healthcare system, the charges were summed across all patients in the sample and then multiplied by 100 (the sample was 1% of all hospital admissions for that year). This yielded an estimate of the healthcare cost of adverse events for New Zealand. The results report the total healthcare cost of adverse events for each patient and for the country. Due to the retrospective nature of the data, it was not possible to identify other economic costs (such as lost wages and household productivity losses). Previous studies found these costs to be significant, representing nearly 48% of the total cost in the Colorado and Utah (Thomas et al, 1999). In order to provide an 3

estimate (albeit crude) of the cost in New Zealand of these other costs, the results report an estimate for the other economic costs in New Zealand based on the assumption that the ratio of healthcare to other costs (.48/.52) is similar in New Zealand as Utah and Colorado. A predictive model was estimated to identify the patient characteristics associated with costly adverse events. As with most data on hospital costs, the distribution of cost data is highly skewed. Thus, the log of cost was used as the dependent variable in the regression analysis. The information available to hospital staff was used as independent variables: Patient demographic information: Patient age, gender and ethnicity were taken from medical records by the medical officers. Socio-economic status: The domicile code showing where the patient lived was taken from medical records. This information was subsequently linked to NZDep96, an area-based index of socio-economic deprivation (Crampton, Salmond and Sutton, 1997) to provide a measure of the relative deprivation of the area in which the patient resided. NZDep96 measures the level of deprivation for each mesh block by combining variables from the 1996 census, including income, access to car, living space, home ownership, employment status, qualifications, support and access to telephones. As such, it is an area level measure of socio-economic status. Illness: The patient s co-morbidity prior to the adverse event was assessed by the medical officers based on the patient s medical history, including reason for admission and overall illness status (see Davis et al, 2001 for a discussion of the reliability and validity of the assessors ratings). Severity of adverse event: Severity was assessed by the medical officers based on the degree and duration of disability resulting from the adverse event. The model was estimated using ordinary least squares regression for the overall sample of adverse events and for the sub-sample of preventable adverse events. 3. Results 3.1 Rate of adverse events, additional hospital days, extra procedures and severity of adverse event 4

As reported elsewhere (Davis et al, 2001), the NZQHS identified 850 adverse events from the review of 6579 medical records. Thus, 12.9% of admissions were associated directly with an adverse event. Of the 848 cases identified that had sufficient information for costing, 37% (310) were determined by the reviewing medical officers to have not been preventable. The remaining 63% (538) were rated as having some degree of preventability. Of these 848 cases, 89% (756) required at least one additional bed day in hospital. The average additional stay (i.e., resulting from the adverse event) spent in the hospital was 9.11 days per patient admission. The average additional stay for patients with preventable adverse events was slightly higher (10.13 days per patient). Forty-three percent (365) of the patients required an additional procedure as a result of the adverse event. The most common additional procedure was surgery, with 204 patients (23%) requiring at least one surgical treatment (some patients required multiple additional surgeries). The rate of surgery was similar for patients with preventable adverse events (22%). As for the long-term effects of the adverse event, 62% were rated as having minimal or no disability (effect lasting less than 1 month) and 19% with moderate disability (lasting 1 to 12 months). Ten percent had some degree of permanent disability, and 4% died as a result of the adverse event (it was not possible to determine the outcome for the remaining 5% of the events). 3.2 Total healthcare costs of adverse events 3.2.1 Cost per patient The total healthcare cost was $10,264 (see Table 2). This included an average of $9,766 in bed day cost (ranging from $0 to over $250,000) and $498 in additional procedures (ranging from $0 to over $12,000). For those events deemed preventable (63% of the total), the average cost per patient was $11,024, ranging from $0 to over $250,000. The average cost of non-preventable adverse events was $8,933, ranging from $0 to just over $125,000. 3.2.2 Total Healthcare Costs The total healthcare cost for all of New Zealand is the sum of the cost of the adverse events in the sample multiplied by 100. Thus, New Zealand spent over $870 5

million as the result of adverse events. Over $590 million was spent on preventable adverse events. 3.2.3 Total Healthcare and other direct costs Assuming that the ratio of other costs (such as wages and household expenditures) to healthcare costs is similar in New Zealand as in the Utah/Colorado study (Thomas et al, 1999), then the total out-of-hospital cost is estimated to be $9,474 per patient (average cost per patient multiplied by the.48/.52). The total cost per patient is estimated at $19,738. For the country as a whole, over $803 million was estimated to be lost as a result of the adverse event in wages and other household expenditures, bringing the total cost of the adverse event to over $1.6 billion (Table 2). 3.3 Predictive model of cost per patient 3.3.1 Patient Characteristics Age: The average age of a patient suffering an adverse event was 51 years. As shown in Table 3, adverse events involving neonates were associated with the highest average cost ($20,531), while those involving 20 to 29 year olds had the lowest average cost ($3,871). For those events deemed preventable, average costs were higher for most age groups with neonates associated with the highest average cost ($26,769). Gender: Women (55%) slightly outnumbered men in the sample (see Table 4) and were associated with less costly adverse events ($8,985 to $11,851). Costs for both women and men were higher when the adverse event was preventable ($10,089 and $12,206, respectively). Co-morbidity: Fifty three percent of the patients had some degree of co-morbidity prior to the adverse event (Table 4). The thirty eight percent whose co-morbidity was rated as moderately serious (MODERATELY ILL) had an average cost of $11,958 (compared with $8,053 for patients with no co-morbidity). Patients rated as having very serious co-morbidity (VERY ILL) were associated with costs of $11,578. All costs rose substantially if the adverse event was deemed preventable.. 6

Ethnicity: People of European descent constitute the largest ethnic group in New Zealand, followed by Maori, those from the Pacific region, and others (primarily Asian). As shown in Table 4, Europeans were associated with the highest healthcare costs ($10,611), followed by Maori ($9,983). Adverse events suffered by Pacific people were the least costly ($7,156). NZDep96: Individuals who reside in the least deprived areas (Decile 1) had an average cost of $9,634, while those in the most deprived areas (Decile 10) had a cost of $11,731 (Table 4). Overall, the average cost of individuals in relatively more deprived areas (Deciles 6 to 10) was slightly higher ($10,657) than those from less deprived areas (Deciles 1 to 5: $9,602). A similar pattern was exhibited for preventable adverse events ($11,602 and $10,104, respectively). 3.3.2 Severity of adverse event As shown in Table 5, sixty two percent of the individuals suffered no or minimal disability as a result of the adverse event. The average cost for these patients ($5,235) was far below the average cost for the 33% of patients who suffered some degree of disability or death as the result of the adverse event. Patients who suffered a moderately severe outcome had an average cost of $15,491. For those with permanent disability, the average cost ($27,415) was over 5 times the average cost of patients with no disability. The average cost of those who died was $14,682. The average costs were greater for those whose adverse event was preventable. 3.3.3 Estimation results The results from the estimation are shown in Table 6. The results shown in columns 1 and 2 are for the entire sample of adverse events, those in columns 3 and 4 look only at preventable adverse events. Looking at the entire sample (columns 1 and 2), the results suggest that the cost of the adverse event increases significantly with age (AGE; t=3.31, p<.001), although events associated with neonates (NEONATES; t=1.94, p<.10) are marginally more costly as well. Costs do not vary with ethnicity (MAORI, PACIFIC and OTHER), deprivation of area (NZDEP), or with gender (FEMALE). The presence of moderately serious co-morbidity (MODERATELY ILL; t=2.97, p<.01) was associated with higher healthcare cost but the significance of very 7

serious co-morbidity (VERY ILL) depended upon the inclusion of other variables. When patient characteristics are viewed in isolation of the severity of the adverse event (column 1), having serious co-morbidity is not a significant predictor of cost. However, when severity of the adverse event is included (MODERATE DISABILITY, PERMANENT DISABILITY and DEATH), having serious comorbidity is marginally associated with higher costs (t=1.76; p<.10). Severity is a strong predictor of cost, with moderate and permanent disabilities being associated with increasing costs (t=3.78, p<.01 and t=4.78, p<.01, respectively) and death with significantly lower costs (t=-2.073, p<.01). These results are consistent with the previous studies examining the determinants of hospital costs (e.g., Graves 2001). The measures of severity are also significant predictors of cost when the analysis is restricted to only those adverse events deemed preventable (columns 3 and 4). Increased costs are associated with age (t=2.32, p<.05), neonates (t=2.37, p<.05), a moderate (t=3.01, p<.01)and serious co-morbidity (t=3.21, p<.01)both without (column 3) and with (column 4) inclusion of the severity of the adverse event. 4. Discussion The current study examined the cost of adverse events associated with admissions to public hospitals in New Zealand. A retrospective review of hospital records identified the additional resources (bed days and procedures) attributable to the adverse events. Charges to international patients were used to estimate the cost of healthcare resources. The results suggest that treating adverse events costs hospitals over $870 million. Given that the total expenditure on care in public institutions in New Zealand was approximately $2.9 billion in 2001 (estimated expenditure on public institutions; New Zealand Ministry of Health, 2000), up to $.30 of every dollar spent by a public institution goes towards treating an adverse event. Thus, adverse events are very costly to the New Zealand healthcare system. The ultimate goal of research examining adverse events is to improve the quality of healthcare and eliminate adverse events. The results from this study suggest that substantial resources (over $590 million) are diverted to treat preventable adverse events. This information is useful when developing interventions aimed at preventing adverse events. For instance, one criterion for deciding upon the merits of a specific intervention is to assess the intervention s expected net benefits (total benefit less total cost). The results from the predictive model highlight populations (elderly or neonate 8

with co-morbidity) where the treatment cost can be reduced should the intervention be a success. When considered in conjunction, the cost of developing and implementing the intervention, the information is useful in identifying the (net) cost of an intervention. Cost information might also serve as a surrogate measure of the pain and discomfort associated with the hospital stay. As shown above, the main drivers of healthcare cost are the additional days spent in hospital. To the extent that pain and suffering that results from the adverse event is correlated with length of stay in the hospital, then the results from the predictive model suggest the characteristics of patients who experience high levels of pain and discomfort as a result of the adverse event. This information can complement injury and disability assessments in identifying the total impact of the adverse event on the patient. The results from the predictive model are broadly consistent with studies examining the characteristics associated with high hospital or treatment costs (e.g., Caro, 1999, Knapp et al, 1993). For instance, Graves (2001) found that the cost of hospital care tends to increase with the age of the patient (although care for neonates tends to be especially costly) and for patients with a moderately severe co-morbidity, and costs tended to be lower if the patient subsequently died. In contrast to the findings in the current study, Graves found that factors such as gender and severe comorbidity were not associated with higher costs. Thus, while the patterns are similar, there are differences between the patient characteristics associated with high hospital costs in general and those specifically relating to adverse events. The method used in this study to estimate healthcare costs (retrospective analysis of hospital cost records) is similar to that employed in previous studies. However, there were differences between the studies in the types of resources that were measured and the method for valuing those resources. In the New York study (Johnson, et al, 1992), the investigators augmented the retrospective analysis (5 years after the event) of the medical records with survey information on 794 patients on health care usage and other direct costs (such as lost wages and lost household production). Market prices were then applied to the resources and lost income. In the Australian study (Wilson et al, 1995), investigators relied on the DRG relating to the sampled admission. The cost of the adverse event was then obtained by multiplying the number of bed days attributable to the adverse event by an average per day cost based on the initial DRG classification. In the Utah and Colorado study (Thomas et al, 9

1999), the investigators used patient records to estimate healthcare utilisation, including inpatient (i.e., bed days) and outpatient (e.g., physical and occupational therapy, home health visits and nursing home care), applying market prices to obtain a final value. Patient demographic information was linked with census records in order to obtain estimates of lost productivity and wages. In the British study (Vincent et al, 2001), investigators considered only additional bed days, applying a per diem cost supplied by the hospital. The current study uses per diem costs by specialty ward (as suggested by Donaldson, 1990) and includes the cost of additional procedures not included in the ward charges. There are a number of limitations of this methodology. First, due to the lack of detail in hospital records, it was not possible in the present study to use a micro or bottom up costing technique (see Jackson, 2000). If patients who have suffered an adverse event require more resources than normal care, then the use of per diem costs may understate the cost of the adverse event. An additional problem with the current study concerns the timing of the adverse event. Although the NZQHS identified the adverse events associated with sampled admissions in 1998, some costs were incurred in 1997 or 1999 (e.g., an adverse event occurring in 1998 but resulting in additional hospital days in 1999 would have been included). Although few in number, their presence slightly inflates the estimated cost of treating adverse events in a given year. A third limitation with the method used in the current study relates to method of assigning a value to the resource (i.e., use of charges to international patients). Unlike the United States, there is no internal competition within New Zealand for a wide range of hospital services. Nor is detailed costing information available for New Zealand s hospital services. The decision to use international prices was made because there is an active market in providing hospital services to international patients in Australasia, with New Zealand hospitals competing with other hospitals (primarily Australian) for patients from the Pacific region and parts of Asia. However, it is unclear how responsive these prices are to international (e.g., exchange rate) or domestic (e.g., government initiatives) events or the degree of competition for services. The total cost would undoubtedly differ if international prices charged by Australian or US hospitals were applied instead. Taken together, these limitations suggest that the results presented here should be seen as an indication of the cost of 10

adverse events but that an alternative design is required for attaining more accurate estimates. To date, no study has conducted a detailed quality of life assessment of adverse events. Rather, the seriousness of the adverse event is typically determined by the extent of permanent disability (including lost future wages and household production costs) that results. This measure ignores the pain and suffering resulting from treatment in the hospital. Because the majority of patients suffer no or minimal disability as a result of the adverse event (and yet may still be subjected to invasive treatments and extended length of stay), disability alone does not provide an accurate assessment of the total effect of the adverse event. Future studies should examine both the decreased quality of life that results from adverse events and, equally important, the characteristics of the hospital environment and medical professional s behaviour that gave rise to the event. This will provide researchers the information needed to develop successful interventions. 11

References Brennan, TA, LL Leape, NM Laird, L Hebert, AR Localio, AG Lawthers, JP Newhouse, PC Weiler and HH Hiatt, Incidence of adverse events and negligence in hospitalised patients: Results of the Harvard Medical Practice Study I, New England Journal of Medicine, 1991, Vol. 324 (6), 370-376. Caro, J, K Huybrechts, and H Kelley, Predicting treatment costs after ischemic stroke on the basis of patient characteristics at presentation and early dysfunction, Stroke, 2001, January, 100-106. Crampton, P, C Salmond and F Sutton, NZDep96: Index of Deprivation. Report No 8, Wellington: Health Services Research Centre, 1997. Davis, P, R Lay-Yee, R Briant, S Schug, A Scott, S Johnson and W Bingley, Adverse events in New Zealand Public Hospitals: Principal findings from a national survey, New Zealand Ministry of Health Occasional Paper Series, No 3, 2001. Donaldson, C., The state of the art of costing health care for economic evaluation, Community Health Studies, 1990, 14 (4), 341-356. Graves, N, Estimating the cost of hospital acquired infection, unpublished PhD dissertation, University of London, 2001. Graves, N and PM Brown, Options for costing hospital based patient care in New Zealand, mimeo, University of Auckland, 2001. Jackson, T., Cost estimates for hospital inpatient care in Australia: Evaluation of alternative sources, Australian and New Zealand Journal of Public Health, 2000, vol. 24 (3), 234-241. Johnson, W, T Brennen, J Newhouse, L Leape, A Lawthers, H Hiatt and P Weiler, The Economic consequences of medical injury: Implications for a No-fault insurance plan, Journal of the American Medical Association, 1992, Vol. 267 (13 May), 18, 2487 2492. Knapp, M and JA Beecham, Reduced listing cost: Examination of an informed short cut in mental health research, Health Economics, 1993, 2: 313-122. Leape, LL, TA Brennan, NM Laird, AG Lawthers, AR Localio, BA Barnes, L Hebert, JP Newhouse, PC Weiler and HH Hiatt, The nature of adverse events in hospitalized patients: Results of the Harvard Medical Practice Study II, New England Journal of Medicine, 1991, Vol. 324 (6), 377-384. Localio, AR, AG Lawthers, TA Brennan, NM Laird, L Hebert, LM Paterson, JP Newhouse, PC Weiler and HH Hiatt, Relation between malpractice claims and adverse events due to negligence: Results of the Harvard Medical Practice Study III, New England Journal of Medicine, 1991, Vol. 325 (4), 245-251. 12

New Zealand Ministry of Health, Health Expenditure Trends in New Zealand: 1980 to 1999, Wellington, New Zealand, 2000. Rigby, K, RB Clark and WB Runciman, Adverse events in health care: Setting priorities based on economic evaluation, Journal of Quality in Clinical Practice,1999, 19, 7-12. Thomas, E, D Studdert, JP Newhouse, BW Zbar, KM Howard, EJ Williams and TA Brennan, Cost of medical injuries in Utah and Colorado, Inquiry, 1999, 36 (Fall), 255-264. Vincent, C, G Neale and M Woloshynowych, Adverse events in British hospitals: preliminary retrospective record review, British Medical Journal, 2001, Vol. 322 (3 March), 517-519. Whynes, D and A Walker, On approximations in treatment costing, Health Economics, 1995, Vol. 4: 31-39. Wilson, R, WB Runciman, RW Gibberd, B Harrison, L Newby and J Hamilton, The Quality in Australian Health Care Study, Medical Journal of Australia, 1995, Vol. 163 (6 November), 458-480. 13

Table 1: Previous Studies examining cost of adverse events Total rate of Adverse events Number of cases examined Preventable Adverse Events Healthcare Cost Region Harvard study (Johnson et al, 1992, JAMA) Colorado and Utah (Thomas et al, 1999, Inquiry) Australia (Wilson et al, 1995, MJA) Britian (Vincent et al, 2001, BMJ) 3.7% 31000 (794 for costs) $3.8 billion New York 2.9% 14321 58% $661 million Colorado and Utah 16.6% 14179 51% $AU 900 million Australia 10.8% 1014 46% 290,000 Greater London 14

Table 2. Total cost of adverse events Healthcare costs: Per Person Healthcare costs for New Zealand Bed Day costs All events $9,766 * Preventable events (63%) (18,093) $10,546 (18,956) Additional procedures $498 (963) $478 (942) Total cost $10,264 (18,443) $11,024 (19,259) All events $827,648,000 $42,739,200 $870,387,200 Preventable events (63%) $567,052,000 $26,039,200 $593,091,200 Healthcare and other costs for New Zealand ( Assumes healthcare costs are 52% of total costs) $1,673,841,624 * All figures are in New Zealand Dollars n = 848 for sample, 848,000 for New Zealand Standard deviations in parentheses 15

Table 3. Average cost of adverse event by age group Age groups All adverse events Preventable Adverse Events % Cost % Cost Neonates 3.2% $20,531 2.4% $26,769 < 5 years 5.2% $8,878 5.0% $6,450 5 to 9 1.9% $4,022 1.7% $6,110 10 to 19 5.9% $9,619 5.0% $13,311 20 to 29 8.8% $3,871 8.5% $4,288 30 to 39 9.6% $6,213 8.9% $7,957 40 to 49 9.4% $10,015 10.0% $8,292 50 to 59 8.0% $16,015 8.7% $16,893 60 to 69 14.5% $8,866 14.7% $9,888 70 to 79 17.1% $11,827 16.7% $12,660 > 80 years 16.4% $12,445 18.4% $12,768 n = 848 n = 538 16

Table 4. Average cost of adverse event by gender, co-morbidity,ethnicity and deprivation All Adverse Events Preventable Adverse Events - % Cost % Cost Gender - Female 55% $8,895 56% $10,089 - Male 45% $11,851 44% $12,206 Co-morbidity - None - Mod. Ill - Very Ill Ethnicity - European - Maori - Pacific - Other NZDEP - Decile 1 - Decile 2 - Decile 3 - Decile 4 - Decile 5 - Decile 6 - Decile 7 - Decile 8 - Decile 9 - Decile 10 47% 38% 15% 74% 15% 4% 7% 5% 6% 8% 7% 10% 12% 10% 14% 14% 13% $8,053 $11,958 $11,578 $10,611 $9,983 $7,156 $8,815 $9,634 $10,100 $6,979 $11,855 $10,117 $11,915 $9,375 $11,227 $9,055 $11,731 44% 46% 10% 74% 15% 4% 7% 6% 5% 9% 7% 10% 12% 10% 14% 13% 13% n = 848 n = 538 $8,065 $13,207 $16,479 $11,806 $9,607 $8,933 $7,590 $11,078 $13,294 $7,990 $10,084 $10,177 $13,962 $9,979 $14,016 $7,059 $12,415 17

Table 5. Average cost by severity of adverse event All adverse events Preventable adverse events Adverse Event Severity % Cost % Cost No or minimal disability 62% $5,235 59% $5,388 Moderate disability 19% $15,491 20% $16,459 Permanent disability 10% $27,415 9% $30,405 Death 4% $14,682 6% $15,897 Unknown outcome 5% $14,581 6% $15,268 n = 848 n = 538 18

Table 6. Regression results: Predicting log of cost of adverse event Dependent variable: Log of cost All adverse events Preventable adverse events Independent variables Coefficients Coefficients Column 1 Column 2 Column 3 Column 4 Intercept 7.137 (.358) 6.901 (.350) 7.233 (.442) 6.908 (.429) Age 0.013 * (.004) 0.013 * (.004) 0.011 * (0.005) 0.013 * (.005) Neonate 1.018 ** (.555) 0.910 ** (.534) 1.833 * (.773) 1.742 * (.741) Female -0.288 (.189) -0.274 (.183) -0.250 (.234) -0.244 (.223) Maori -0.315 (.274) -0.323 (.265) -0.256 (.351) -0.294 (.334) Pacific -0.014 (.512) -0.016 (.495) -0.053 (.603) -0.008 (.577) Other ethnicity -0.413 (.415) -0.554 (.402) -1.010 * (.476) -1.135 * (.456) Moderately Ill 0.612 * (.206) 0.632 * (.200) 0.739 * (.245) 0.819 * (.236) Very ill 0.440 (.277) 0.472 ** (.268) 1.310 * (.408) 1.317 * (.390) NZDEP -0.015 (.036) -0.019 (.035) -0.027 (.045) -0.025 (.043) Adverse event severity 0.903 * 0.941 * - Moderate disability (.239) (.283) - Permanent disability 1.478 * (.309) 1.767* (.401) - Death -2.073 * (.449-2.063 * (.485) R 2 =.03 R 2 =.10 R 2 =.06 R 2 =.14 n = 848 n = 839 N = 538 n = 533 Standard errors shown in parentheses * Significant at.05 ** Significant at.10 Dummy variables were used to indicate whether the patient was a neonate, gender (= 1 if FEMALE), ethnicity, degree of co-morbidity (MODERATELY ILL or VERY ILL) and severity of adverse event (MODERATE DISABILITY, PERMANENT DISABILITY and DEATH). 19