Article Text

other Versions

Download PDFPDF
Patients' reports of adverse events: a data linkage study of Australian adults aged 45 years and over
  1. Merrilyn Margaret Walton1,
  2. Reema Harrison1,
  3. Patrick Kelly1,
  4. Jennifer Smith-Merry2,
  5. Elizabeth Manias3,
  6. Christine Jorm4,
  7. Rick Iedema5
  1. 1School of Public Health, University of Sydney, Sydney, New South Wales, Australia
  2. 2Faculty of Health Sciences, University of Sydney, Sydney, New South Wales, Australia
  3. 3School of Nursing and Midwifery, Deakin University, Burwood, Victoria, Australia
  4. 4Sydney University, Sydney Medical School, Sydney, New South Wales, Australia
  5. 5Monash University, Clayton, Victoria, Australia
  1. Correspondence to Dr Reema Harrison, School of Public Health and Community Medicine, University of New South Wales, Sydney, NSW 2052, Australia; reema.harrison{at}


Background Understanding a patient's hospital experience is fundamental to improving health services and policy, yet, little is known about their experiences of adverse events (AEs). This study redresses this deficit by investigating the experiences of patients in New South Wales hospitals who suffered an AE.

Methods Data linkage was used to identify a random sample of 20 000 participants in the 45 and Up Cohort Study, out of 267 153 adults aged 45 years and over, who had been hospitalised in the prior 6 months. A cross-sectional survey was administered to these patients to capture their experiences, including whether they had an AE and received honest communication about it.

Results Of the 18 993 eligible participants, 7661 completed surveys were received (40% response rate) and 474 (7%) reported having an AE. Most AEs related to clinical processes and procedures (33%), or medications and intravenous fluids (21%). Country of birth and admission through emergency were significant predictors of the occurrence of an event. An earlier admission in the prior 6 months or a transfer to another healthcare facility was also associated with more AEs. Of those who suffered an AE, 58% reported serious or moderate effects.

Conclusions Given the exclusions in our sample population (under 45 years), the AE rate reported by patients of 7% is similar to the approximately 10% rate reported in the general population by retrospective medical record reviews. AE data that include patient experience may provide contextual information currently missing. Capturing and using patient experience data more effectively is critical; there may be opportunities for applying co-design methodology to improve the management of AEs and be more responsive to patients' concerns.

  • Patient safety
  • Patient satisfaction
  • Patient-centred care
  • Adverse events, epidemiology and detection

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Concerted efforts to improve the safety and quality of healthcare by governments worldwide have not resulted in a measurable decrease in adverse events (AEs).1 Estimates of the rate and type of harm vary, but it is generally accepted that about 10% of hospital admissions result in some form of harm.2–4 AE data are typically collected using retrospective medical record review and incident reporting, but these activities are compromised by inaccurate or incomplete medical records and under-reporting in voluntary reporting systems.5–7 These methods also fail to capture the patient experience of AEs. The continued struggle by health professionals to reduce harm to patients obliges us to understand why such little progress has been made. Do we understand all the factors contributing to iatrogenic harm?

The routine collection of patient experience data at frequent intervals and at a local level by Australian hospitals is rare; instead, hospitals may rely upon patient ‘satisfaction’ surveys to fill that void. Patient satisfaction surveys usually include rating questions about how satisfied patients feel about their care as a ‘customer’ of the service rather than asking them ‘what happened?’.8 Berwick notes that needed service improvements can only be deduced from specific data about the nature of events as they occur; customer satisfaction ratings do not provide the in-depth evidence required.9

While patients may take a broader view of AEs than health providers,10–12 studies comparing patient accounts of medical errors to hospital records show that patients can accurately identify AEs8 and that many patient-identified events are additional to those recorded or reported in systems for capturing AEs.12–14 A voluntary online survey involving 696 self-selected patients who experienced an AE, showed harm was mostly associated with diagnostic and therapeutic errors, surgical and procedural complications, hospital-associated infections and medication errors.15 That study, like many others was limited due to the bias inherent in having patients self-select to participate or bias created when health services select the patients to participate in such research. Our mixed-methods large-scale control study of the patient experience of AEs avoids many of the problems associated with previous patient experience research.



This mixed-methods study involved data collection via survey from a large research cohort and data linkage between the Admitted Patient Data Collection (APDC), the Register of Births, Deaths and Marriages, and the 45 and Up Study databases. Details of the methods and sampling are reported elsewhere.16


The sample was obtained from the 45 and Up Study cohort, which involves 267 000 people in the state of NSW, Australia aged 45 years and over.17


Eligible participants were a random sample of 20 000 individuals within the 45 and Up Study cohort who were admitted to any hospital in NSW between January and June 2014. Exclusion criteria for participants were those for whom the postal survey was returned to sender, who were reported as deceased, or who responded to say the data linkage was not correct regarding their hospitalisation. Participants, including those with multiple admissions during the study period, were asked to report on their most recent hospitalisation.

While generally representative of the Australian population, the 45 and Up Study sample is not representative with respect to individuals from culturally and linguistically diverse (CALD) backgrounds.17 For example, while only 25% of the 45 and Up Study were born outside of Australia, the 2006 census data put this figure at 39% for those aged 45 years and over in NSW.18


The Centre for Health Record Linkage (CHeReL) is a NSW Ministry of Health agency. CHeReL linked data from the APDC with the 45 and Up Study Database to identify participants who were hospitalised in the study period using admission data. The full data collection procedure is reported elsewhere.16

There were five parts of the survey, which was developed by the research team, ranging from part A to part E (see online supplementary file 1). Participants who did not experience an AE only completed parts A and B of the survey, which involved details about participants' hospital stay and their experience of this stay, respectively. Participants who experienced an AE completed a further three parts (C, D and E) in which they answered questions about the AE experienced, the open disclosure process and whether they made a complaint, respectively. The project aimed to capture data about any incident that the patient believed to be an AE; unintended or unnecessary harm caused by the healthcare they received. The definition used was ‘an event or circumstance during healthcare caused by the hospital which could have or did result in unintended or unnecessary harm to you’. This definition was based on the most commonly cited definitions of AE but modified to be appropriate to a lay audience with the key aspects retained of: (1) unintended or unnecessary, (2) harm resulting and (3) caused by the healthcare provider (in this case, the hospital).19–21

supplementary data


Analyses were conducted using Stata V.13.0. Descriptive statistics were used to describe the frequency and type of AEs reported. Following descriptive analysis, a χ2 test was used to determine if patients who experienced AEs differed significantly from patients who did not experience AEs based on the following patient and admission characteristics: age, gender, marital status, country of birth, education level, whether they had English as a first language, emergency status, local health district area; disease(s) type, care type, hospital type, length of stay, transfer while admitted and whether they had a weekend admission (see table 1 for the categories for each item).

Table 1

Participant characteristics of those who did and did not have an adverse event

We sought to address the limited representation of CALD participants by analysing a subset of the surveys we receive from CALD participants to compare the experiences of those CALD participants who experienced an AE with those who did not experience an AE. We also compared the CALD participants in the 45 and Up cohort to the non-CALD members of the cohort to see if there were specific differences or variations in the characteristics associated with the AE, but there were no significant differences.

Thematic analysis was conducted of the free-text response box of the survey, which asked patients what happened to them. The text was coded according to the WHO International Classification for Patient Safety categories.19 This classification system was used to ensure that foreseeable and known complications of conditions and treatments were separately identified and described. We noted that a significant number of patient incidents fell outside the WHO classification system. These events that fell outside of the classification system were also identified, grouped and labelled resulting in four patient-defined categories (healthcare problem admitted for; problem during/after care; readmission; costs associated with care) that were additional to the classification framework. Reliability was ensured through individual categorisation of all responses by two members of the research team (RH and JS-M). Differences in categorisation were resolved through discussion between the two researchers.


Preliminary analysis

Of the 20 000 potential participants from the 45 and Up Study who were invited to participate, 18 993 were eligible to take part. Participants were considered ineligible if the postal survey received was returned to the sender (640), were reported as deceased (189), or responded to say the data linkage was not correct and that they had not been in hospital (178). Completed surveys were received from 7663 of those eligible to participate (40% response rate).

χ2 tests confirmed there were no significant differences between responders and non-responders in age group (p=0.95), gender (p=0.49), the proportion for whom English is not their only language (p=0.63), local government area (p=0.84) or level of education (p=0.93). Participants were evenly distributed in terms of gender, public versus private hospital admissions, whether they attended an acute hospital or not and across the age range from 50 years to 110 years. Most participants spoke English as their first language (91%), did not have a weekend admission (91%), were in an acute setting (98%) and had a planned admission (72%). Table 1 presents a summary of the sample characteristics.

Seven per cent of participants reported that they experienced an AE during their care, defined in our survey as ‘An event or circumstance during healthcare caused by the hospital which could have or did result in unintended or unnecessary harm to you’. Two factors were identified as significantly associated with participants' reporting an AE: emergency status and country of birth. Being born in South America and those who had emergency admissions were independent predictors of reporting having AEs. AEs were reported as 1.4% higher in emergency than planned admissions (95% CI 0.1% to 2.7%). Higher proportions of those who had a transfer during their admission or who had a previous admission in the 6 months prior to the study period also reported having an AE; however, in both cases these were not statistically significant (p=0.07). None of the other explanatory factors assessed were significant predictors of the occurrence of an AE (see tables 1 and 2).

Table 2

Admission and care characteristics of those who did and did not have an adverse event

Types of AEs reported

Patients reported a range of AEs, with 474 cases. These cases accounted for 633 incidents, which were classified using the WHO patient safety taxonomy.19 The WHO incident categories relevant to this study are listed in table 3 under incident type (see WHO, 2009 for full list of categories). Most incidents identified by patients related to clinical processes, procedures, medications or intravenous fluids (see table 3 for a breakdown). These categories encompassed a broad range of events, which commonly included delayed, missed or incorrect medication, wrong site injections and problems in the insertion of catheters and needles. Of the reported events, 272 (58%) were described as having (patient-defined) moderately severe or severe effects, with pain being the most common. Only 76 (16%) had no resultant effect on the patient. Ninety-six events were identified that did not fit the WHO categorisation criteria. In 86 cases, patients did not provide sufficient detail in their description to categorise the event with 52 patients simply giving the reason they were in hospital and 34 patients just reporting that a problem had occurred during the care process. In 20 of these cases this was because the event was something that resulted from the AE—in 10 cases a readmission, in 9 cases patients were told there was a complication and in 1 case the patient identified the cost associated with the AE as the harm they incurred.

Table 3

Frequency of incidents relating to each WHO incident category


Respondents in this study reported that AEs occurred in 7% of hospital admissions in NSW during the study period. This is likely a conservative estimate given that deceased patients were excluded and those who were extremely sick were likely to be non-responders. Importantly, while most of the AEs reported (503/633 events; 79%) were within the boundaries of the clinical definition of an AE, patients placed importance on and reported the non-clinical events arising before and after the harm occurred (130/633 events; 21%), including communications with staff, financial burden, readmission to hospital. In this way, our findings offer an alternative profile of AEs that cannot be directly compared with the estimates derived from record review or incident reporting studies.

Characteristics associated with AEs

AEs were more commonly reported among those who had an emergency admission, reflecting existing research literature.22–24 This is likely due to the high volume, high pressure care associated with emergency admissions. More AEs were also reported by those who had experienced a transfer during their care process; although not a significant finding, this is important when considering risk factors for AEs. Transfers between healthcare professionals, healthcare services and healthcare institutions have long been recognised as posing threats to patient safety, particularly when clinical handover is poor.25 Information sharing between teams and services, communication and clinicians' understanding of the necessary information to transfer have all been identified as vital for effective handover. Making up for the gaps in information provision between providers and services has been identified as one of the contributions that patients make to their care in the research literature.26

Country of origin also appeared to be an important factor, with those from South America identified as reporting significantly more AEs. The research evidence does not identify South American patients to be particularly vulnerable to AEs. However, we do know that older adults who are CALD have been identified as more susceptible to experiencing some types of AEs, especially when they have lower English language proficiency.27 This unexpected finding may be in part explained by the challenge of identifying CALD participants from our study data and the linked data sets. CALD patients were identified from their medical records as those who spoke a non-English language as their first language at home. While this group were not significantly more likely to report an AE, the variable used did not provide information about English proficiency and may not have been sufficiently sensitive. It is possible that while many of the study participants were non-native English speakers, the South American participants in this particular study may have had a lower English proficiency that led to a greater vulnerability to AEs when in hospital.

Our data suggest that no one type of patient is more susceptible to an AE; a patient irrespective of age and/or comorbidities has the same chance of suffering an AE as others who are admitted to hospital. The results also showed no significant difference between the number of AEs reported by those admitted to public or private hospitals, with weekend admissions or who had longer stays, which have previously been identified as predictors of AEs.28 It is possible that while there were no significant differences between responders and non-responders in their demographic characteristics, those who have the most complex conditions, comorbidities and suffer the greatest detrimental health effects in relation to older age are also those who were less able to respond to the study survey. In addition, it may be that patients who are participants in the 45 and Up Study are generally in better health than those who have chosen not to be study participants.

Defining AEs

One of the key findings in our study is that there are differences between what a health service considers an AE compared with the reports of patients. This study provided patient respondents with the opportunity to define for themselves the boundaries of an AE caused by their medical management that resulted in harm to them. Most of the events reported (79%) could be categorised to the WHO patient safety taxonomy, which provided an existing framework for classifying AEs based on health services definitions.19 The taxonomy did not, however, reflect some of the issues identified as AEs by patients; notably those that related to issues arising before or after the clinical harm. Patients identified complications, extended stays and financial implications of additional care as events in themselves. This finding highlights some of the complexities in determining an AE; for example, if a complication results from an inherent risk of a procedure but these were not discussed with the patient, it could be considered an AE by the patient but a complication by the clinician. Such experiences are generally considered a result of an AE by health services or are not therefore included in the WHO taxonomy.

Other studies of patients-reported AEs highlight the wider scope in definition of an AE by patients often linked to the concept of harm.29 The narrow definition of an AE, while helpful for health services in terms of AE surveillance, potentially ignores some of the most pertinent issues surrounding AEs for patients. If healthcare is to be patient-centred, our findings suggest that healthcare providers should place more emphasis on addressing the overall impact of the AE—both the causes and the outcomes of the events—as these are clearly important to patients.

Patient involvement

The findings demonstrate areas in which patient-reported AE information may be additional to that of health professional reports and helpful for improving care quality in relation to a range of different types of incidents.8 ,12–15 For example, by adding context to event information, identifying quality of care issues and problems related to treatment, especially medications, during or after the care process.8 After-event surveys of patient experience, including ours, may not be sufficiently refined to capture these nuanced data. Yet the routine capture of qualitative data via ‘exit interviews’ for patients raises additional challenges in resourcing and capturing comparable data on scale. Applying experience-based co-design (EBCD) methods might be one approach to more effectively enhance the management of AEs and increased responsiveness to patients' needs and preferences.30 The EBCD framework is one of many approaches to co-design, drawing upon the lived experiences of patients, families and health professionals to collaborate and redesign healthcare services.31 It may provide a minimally onerous sustainable strategy to capture and use in-depth patient experience data, focusing on particular types of AEs or areas within a particular health service.


Due to sensitivity surrounding death we did not capture data about the experiences of patients who had died. We linked our study data to death data which indicated that 189 people had died in the study period. Some of those excluded may have had an AE and those AEs resulting in death during this period cold also therefore have been excluded from our data. While we confirmed no demographic differences between responders and non-responders, there may have been different experiences among the non-responders that have not been captured here.


Combining patient experience of AEs with hospital measures may capture patients' perspectives, particularly regarding quality concerns. These results indicate that patient-reported AEs may provide missing evidence that can be used in efforts to reduce AEs. Patient-centred care requires patient safety to be considered as part of providing high quality care rather than as a distinct aspect. In order to do this, new tools that more effectively capture AE experience and allow the inclusion of patients' concerns regarding quality of their care are needed.



  • Contributors Each of the listed authors made substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data; drafting the article or revising it critically for important intellectual content; and to the final approval of the version to be published.

  • Funding National Health and Medical Research Council (1049703).

  • Competing interests None declared.

  • Ethics approval Approval granted from the NSW Population Health Research Ethics Committee.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Our data will be accessible via the 45 and Up Study and their secure storage systems, which are accessible for a fee to the Sax Institute on application.