The purpose of this study was to determine the rate of self-reported errors in Canada compared with other countries, and to identify risk factors for medical error. In 2007, the Commonwealth Fund surveyed a sample of adults in seven industrialized nations, including Canada. Data from this source were used to perform a bivariate analysis comparing those individuals who reported having experienced a medical error with those who did not, followed by a logistic regression model to delineate the relationship between medical error and several explanatory variables.

Overall, 11,910 respondents from seven countries were included in the analysis. The rate of self-reported medical error ranged from 12 to 20% in the seven nations. Approximately one in six Canadians reported having experienced at least one error in the past two years, which translates to 4.2 million adult Canadians. Several variables were found to have a statistically significant relationship to self-reported medical errors in the final regression model, including high prescription drug use, the presence of a chronic condition, a lack of physician time with the patient, age under 65, a lack of patient involvement in care, perceived inadequate nursing staffing and an absence of a regular doctor. Identification of several patient, provider and system characteristics associated with self-reported medical error should aid in the development of strategies to address this problem by healthcare decision-makers and clinicians.


Canada, like other industrialized countries, aims to provide safe healthcare to its citizens that meets the fiscal responsibilities of delivery in an efficient manner. This is evident from many initiatives across the world aimed at improving patient safety, such as the Institute for Healthcare Improvement's 100,000 Lives Campaign in the United States and the Safer Healthcare Now! initiative in Canada. A key component of improving patient safety is the prevention and management of medical error. Successful management of such errors and changes aimed at reducing the likelihood of their occurrence can have a significant impact on improving patient safety and quality of care, thereby resulting in fiscal benefits to the system.

There is substantial evidence suggesting that medical errors are a leading cause of death and injury (Kohn et al. 1999). According to the Canadian Adverse Events Study, 7.5% of patients admitted to acute care hospitals in Canada during 2000 experienced an adverse event, with 36.9% of those errors being highly preventable (Baker et al. 2004). Such events ranged from the wrong dose or type of medication to having a foreign object left inside the body following a surgery, with an associated 1.1 million extra days in hospital attributed to these events (Gagnon 2004). This accounts for 5.4% of the total hospital days in Canada (Canadian Institute for Health Information 2004). In their review of the literature, Baker and colleagues (2007) concluded that the situation in Canada is roughly equivalent to that in many other industrialized nations.

Medical errors can result from issues related to patients, providers or the healthcare system. The main objectives of this study were to determine (1) the self-reported rates of medical, laboratory and medication errors in Canada, Australia, Germany, the Netherlands, New Zealand, the United Kingdom and the United States, and (2) the patient, provider and system-related risk factors for self-reported medical error.


The Commonwealth Fund's 2007 International Health Policy Survey in Seven Countries, conducted in Australia, Canada, Germany, the Netherlands, New Zealand, the United Kingdom and the United States, was the primary data source for this paper. The data used in this survey were collected by Harris Interactive Inc., and its collection and sample procedures are detailed in full elsewhere (Schoen et al. 2007). In short, surveys were conducted by telephone in 2007 by Harris Surveys and country affiliates, with an average interview time of 17 minutes. We received permission from the Commonwealth Fund to use the raw survey data for the purpose of this paper.

A bivariate analysis of the survey data was performed using a chi-square test of significance with a = .01 comparing those respondents who reported having experienced at least one medical error in the past two years with those who did not report having experienced an error. All the variables included in the bivariate analysis were hypothesized to be possible explanatory variables (risk factors) for medical error, based on previous literature. These included factors related to patient demographics, access to care, coordination of care and the patient-provider relationship. In comparing patients experiencing and not experiencing an error, chi-square was used to test for any significant difference in the variables for the purpose of identifying variables to be included in logistic regression models. If the chi-square value for the explanatory variable was significantly associated with an error (p < .01), it was then included in the regression model. Logistic regression was carried out to determine if any statistically significant exploratory variables from the bivariate analysis were predictors of medical error. For the risk factors of self-reported medical error in the final regression model, the parameter estimate, standard error and chi-square probability were all calculated. The odds ratios and 95% confidence intervals were also determined for each risk factor to provide an estimate of the relative risk of having a self-reported medical error given the presence of the risk factor. The goodness of fit of the final model was assessed, as was the possible presence of multicollinearity. All data analysis was performed using SPSS Version 16.0 (SPSS Inc., Chicago, Illinois).


The final data set for this paper consisted of adults aged 18 and older in seven different countries: Australia, Canada, Germany, the Netherlands, New Zealand, the United Kingdom and the United States, totalling 11,910 survey respondents. Demographic characteristics of this group are listed in Table 1.


Table 1. Demographics of the survey respondents (N = 11,910)
Country Total Respondents Male Female 65 Years of Age Chronic Illness*
n n % n % % %
Australia 1,009 365 36.2 644 63.8 21.5 60.7
Canada 3,003 1,370 45.6 1,633 54.4 19.5 50.0
Germany 1,407 660 46.9 747 53.1 16.2 44.9
Netherlands 1,557 635 40.8 922 59.2 27.5 68.3
New Zealand 1,000 390 39.0 610 61.0 17.6 53.6
United Kingdom 1,434 557 38.8 877 61.2 28.3 50.4
United States 2,500 942 37.7 1,558 62.3 23.1 60.9
All 11,910 4,919 40.7 6,991 59.3 22.0 55.5
*Defined as one of the following conditions: arthritis, heart disease including heart attack, diabetes, asthma, chronic obstructive pulmonary disease, high blood pressure or hypertension, depression, cancer, chronic pain or a mood disorder other than depression such as bipolar or mania.


Table 2 displays the self-reported medical, medication and laboratory errors in each of the seven countries. Overall, the percentage of respondents who reported having experienced at least one error ranged from a low of 12% in Germany to high of 20% in Australia and the United States. Canadians had the third highest self-reported error rate at 17%. This translates to over 4.2 million adult Canadians (17% of 24.7 million adult Canadians in the 2006 census [Statistics Canada 2006]) who experienced at least one error during this time frame.


Table 2. Medical, medication and laboratory errors in seven countries*
Medical, medication errors In past 2 years (percent yes):
Been given the wrong medication or wrong dose? 8%c 6% 5% g 6% 6% 6% 7%
Had a time when you thought a medical mistake was made in your treatment or care? 11% b,c,d,f 7% d,g 6% g 5% e,f 8% f 5% g 9%
Either medical or medication error 15% b,c,d,f 10% g 9% g 9% g 11% 9% g 13%
Laboratory errors
Had blood tests, radiographs or other tests in past 2 years (percent yes) 82% b,d,f 79% c,d,f 81% d,f 72% e,f,g 79% f 62% g 79%
Been given incorrect results for diagnostic or laboratory test 5% c,d 4% c,d,f 2% g 2% g 2% g 2% g 5%
Experienced delays in being notified about abnormal results 7% c,g 9% c,e 2% d,e,f,g 7% g 7% g 8% 11%
Either laboratory or diagnostic error 11% c 12% c,d,e 4% d,e,f,g 8% g 9% g 10% g 14%
Combined medical, medication, or laboratory errors
Any medical, medication or laboratory error (percent yes) 20% c,d,f 17% c,f,g 12% c,g 14% g 16% f,g 13% g 20%
Any error, number of doctors seen in past year
One 15% h 13% h 7% h 9% h 13% h 9% h 17% h
Three or more 27% 30% 16% 27% 34% 29% 33%
Any error, number of chronic conditions
One 19% h 17% h 14% 15% h 20% 13% h 21% h
Two or more 26% 28% 16% 25% 22% 24% 32%
Among adults with chronic condition, any error for those with and without a medical home
Has medical home 18% h 17% h 11% h 17% h 15% h 11% h 21% h
No medical home 30% 29% 19% 22% 30% 26% 34%
AUS = Australia; CAN = Canada; GER = Germany; NETH = The Netherlands; NZ = New Zealand; UK = United Kingdom; USA = United States of America.
* Reading from left to right starting with Australia, the letter indicates significant differences with countries to the right (p < .05), as indicated: b = different from Canada; c = different from Germany; d = different from The Netherlands; e = different from New Zealand; f = different from the United Kingdom; g = different from the United States; h = indicates difference within country (p < .05).
Source: Commonwealth Fund International Health Policy Survey, 2007. Table and notes reproduced with permission from Schoen et al. (2007).


In the bivariate analysis (Table 3), a majority of the hypothesized risk factors for error occurred more frequently in those who experienced an error (p < .01). This included variables related to patient demographics, access to care, coordination of care and the patient-provider relationship. In fact, only one of our hypothesized risk factors (patients reported not receiving written medical information) did not occur more frequently in the patients who experienced an error, although it neared statistical significance (p = .069). It should be noted that for one variable (age), the relationship to medical error was opposite what we initially hypothesized, as we observed that younger (under 65) respondents were more likely to report having experienced an error.


Table 3. Bivariate analysis of respondents with, and without, any medical error
Variable Patients with Error (n = 1,938)* Patients without Error (n = 7,652)* p Value (asymp. sig. [2-sided])
n % n %  
Male 735 37.9 3,092 40.4  
Female 1,203 62.1 4,560 59.6 .046 c
Chronic illness
No 567 29.3 3,072 40.1  
Yes (at least one) 1,371 70.7 4,580 59.9 .000 b
Poor physician explanation Rarely or never to, "When you need care or treatment, how often does your general practitioner/regular doctor/the doctor- explain things in a way you can understand?" 87 4.7 104 1.4 .000 b
Poor physician knowledge of medical history Rarely or never to, "When you need care or treatment, how often does your general practitioner/regular doctor/the doctor- know important information about your medical history?" 102 5.6 164 2.3 .000 b
Lack of time with physician Rarely or never to, "When you need care or treatment, how often does your general practitioner/regular doctor/the doctor- spend enough time with you?" 174 9.5 248 3.4 .000 b
18-64 1,534 79.2 5,772 75.4  
65+ 404 20.8 1,880 24.6 .001 a
Medication use
0 medications 514 26.6 2,625 34.5  
4+ medications 639 33.1 1,776 23.4 .000 b
Patient involvement in care Yes, definitely to, "Were you involved as much as you wanted to be in decisions about your care and treatment?" 377 50.9 1,151 67.8 .000 b
Enough hospital nurses Yes to, "In your opinion, were there always or nearly always enough nurses to care for you during your hospital stay?" 497 66.7 1,405 82.5 .000 b
Written medical information Yes to, "Did you receive written information on what to do when you returned home and what symptoms to watch for?" 483 64.9 1,168 68.7 .069 c
Regular physician? Yes to, "Regular doctor?" 1,736 89.6 7,083 92.6 .000 b
a = significant difference in variable between respondents with, and without, any error, p < .01 (two sided); b = significant difference in variable between respondents with, and without, any error, p < .001 (two sided); c = no significant difference.
* Of the total number of survey respondents, 9,590 qualified to report on error survey questions. Qualification for these questions was a hospital admission within the past 2 years at the time the survey was administered. A chi-square test of significance with a = .01 was used.


Through the use of logistic regression with self-reported medical error as the dependent variable and nine hypothesized risk factors from the bivariate analysis as independent variables, a seven-variable risk model was produced. Table 4 contains the final overall model with the seven risk factors for self-reported medical error: presence of a chronic condition, a lack of physician time with the patient, age under 65, use of four or more medications, a lack of personal involvement in care, perceived inadequate nursing staffing and the absence of a regular physician. All of these variables were our initial hypothesized risk factors for error, except age less than 65.


Table 4. Logistic regression model results with risk factors for medical error
Variable Parameter Estimate (b) SE x2 Probability OR 95% CI
Lack of patient involvement in care 0.753 0.138 0.000 2.124 1.620-2.784
Perceived inadequate nursing staffing 0.560 0.119 0.000 1.751 1.388-2.210
Absence of a regular doctor 0.679 0.233 0.004 1.972 1.248-3.312
Use of 4+ medications 0.264 0.118 0.026 1.302 1.033-1.642
Age 65+ -0.493 0.130 0.000 0.611 0.474-0.788
Lack of physician time with patient 0.508 0.154 0.001 1.662 1.228-2.249
Presence of a chronic condition 0.504 0.142 0.000 1.655 1.252-2.187
Equation constant -3.864 0.416 0.000 - - -
CI = confidence interval; OR = odds ratio; SE = standard error.



This study confirms that self-reported medical error is a common problem in Canada and other comparable industrialized nations. While the rate of self-reported error ranged from 12 to 20%, Australia, the United States and Canada ranked the worst among the seven countries, which closely follows a 2005 survey of patient perceptions of safety in which these same three countries were the worst of six compared countries (Davis et al. 2006).

Our final prediction model for self-reported medical error appears to be rigorous. Several indices were used to assess model fit. A test of the full model versus a model with intercept only was statistically significant (chi-square = 161.04, p < .001), and the model was able to correctly classify 79% of cases. Also, the Hosmer-Lemeshow Goodness-of-Fit test was not significant, indicating the data fit the model well (chi-square = 4.7, p = 0.699). Overall, the correlations (pairwise relationships) between the predictor variables are quite low, ranging from 0.05 to 0.23. The very low risk of multicollinearity among predictors was also confirmed by visual examination of predictor scatterplots.

The risk factors in our final prediction model include system, patient and provider characteristics. All of these were hypothesized risk factors with one exception - age under 65 (see Table 4). Our findings may be explained, in part, by the cultural differences between the two cohorts, whereby those in the older cohort are more familiar with a paternalistic style of medicine and, as such, are less likely to question outcomes from their healthcare experiences. Another variable to consider is education of the two groups. There has been a trend toward increased education over the years; we may conclude that formal education, which is also reflected by socioeconomic status, is less for the majority of the older cohort, which may influence their propensity not to report errors, perhaps from not recognizing that an error occurred or from being intimidated by a system that appears more paternalistic to such groups.

Based on the results of our study, we have two primary recommendations for clinicians and healthcare policy makers. First, we suggest the creation of a proactive risk identification system for medical error based on the model generated in this study. When patients present at a physician office, physicians could assess risk according to the three types of characteristics in our model: risk related to patient characteristics, physician characteristics and the system. Moreover, a similar process could be carried out during patients' admission to hospital, in that their chart could be tagged with an error risk level of high, moderate or low depending on the circumstances. Second, both healthcare organizations such a regional health authorities and individual clinicians should strive to encourage patient involvement in care. In our model, a lack of perceived involvement in decision-making was the risk factor with the highest odds ratio associated with self-reported medical error.

The primary limitation of this study is the self-reported nature of the survey data. This may produce variations in results due to differences in interpreting the definition of error by the survey respondents, especially across multiple countries. An attempt to draw generalizations and make comparisons from system-wide data at the country level can also lead to ecological fallacy due to variations at each level. However, there is consistency of self-reported error rates across the seven countries (see Table 2), which suggests that the self-reported data are valid.


This study has demonstrated that medical error is a commonly occurring problem from the perspective of patients in seven industrialized countries. The risk factors for self-reported medical error that have been identified in this study should aid clinicians and healthcare policy makers in the design and implementation of targeted strategies to address this issue. By proactively identifying patient-, provider- and system-related risk factors for medical error, the opportunity for improving the safety of Canada's healthcare system could be greatly enhanced.

About the Author

Joshua O'Hagan, MHA, is a medical student at the Faculty of Medicine, Memorial University, St. John's, Newfoundland and Labrador.

Neil J. MacKinnon, PhD, FCSHP, is the associate director for research and a professor in the College of Pharmacy, and a professor in the School of Health Administration and Department of Community Health and Epidemiology, Faculty of Medicine at Dalhousie University, Halifax, Nova Scotia. You can contact him by phone at 902-494-6379, by fax at 902-494-1396 or by e-mail at neil.mackinnon@dal.ca.

David Persaud, PhD, is an associate professor in the School of Health Administration, Dalhousie University.

Holly Etchegary, PhD, is a research scientist with Eastern Health, St. John's, Newfoundland. She is a social psychologist and applied health researcher with primary research interests in psychosocial genetics and knowledge transfer in healthcare.


We would like to acknowledge Michelle Doty, Robin Osborn, Cathy Schoen and Katie Shea at the Commonwealth Fund (New York, New York) for the provision of the data used in this paper. We would also like to acknowledge Sean Higgins and Jennifer Isenor for their contributions in reviewing this paper.


Baker, G.R., L. Jeffs, M. Law, P.G. Norton. 2007. "Improving the Safety and Quality of Health Care in Canada." In N.J. MacKinnon, ed., Safe and Effective. The Eight Essential Elements of an Optimal Medication-Use System. Ottawa, ON: Canadian Pharmacists Association.

Baker G.R., P.G. Norton, V. Flintoft, R. Blais, A. Brown, J. Cox, E. Etchells, W.A. Ghali, P. Hebert, S.R. Majumdar, M. O'Beirne, L. Palacios-Derflingher, R.J.Reid, S. Sheps and R. Tamblyn. 2004. "The Canadian Adverse Events Study: The Incidence of Adverse Events among Hospital Patients in Canada." Canadian Medical Association Journal 170(11): 1678-86.

Canadian Institute for Health Information. 2004. Healthcare in Canada 2004: Report Highlights, Part A; Focus on Safe Care. Ottawa, ON: Author. Retrieved January 10, 2009. < http://secure.cihi.ca/cihiweb/dispPage.jsp? cw_page=AR43_2004highlight_e > .

Davis, K., C. Schoen, S.C. Schoebaum. 2006. Mirror, Mirror on the

Wall: An Update on the Quality of American Health Care through the Patient's Lens. New York, NY: The Commonwealth Fund.

Gagnon, L. 2004. "Medical Error Affects Nearly 25% of Canadians." Canadian Medical Association Journal 171(2): 123.

Kohn, L.T, J.M. Corrigan and M.S. Donaldson, eds. 1999. To Err Is Human: Building a Safer Health System. Washington, DC: National Academy Press, Institute of Medicine.

Schoen, C., R. Osborn, M. Doty, M. Bishop, J. Peugh and N. Murukutia. 2007. "Toward Higher Performance Health Systems: Adults' Healthcare Experiences in Seven Countries, 2007." Health Affairs 26(6): w717-34.

Statistics Canada. 2006. Age (123) and Sex (3) for the Population of Canada, Provinces, Territories, Census Divisions and Census Subdivisions, 2006 Census - 100% Data. Ottawa, ON: Statistics Canada. Retrieved January 17, 2009. < http://www12.statcan.ca/english/census06 > .