Objective To discuss the characteristics of incidents reported to the Medical Safety in Community Practice (MSCP) safety learning system.
Methods Members of family physician offices in the Alberta Health Services—Calgary zone, confidentially reported patient safety incidents via web or fax from September 2007 to August 2010. The incident reporting form contained both open-ended and closed questions. Incidents were reviewed for their characteristics.
Results A total of 19 family practices participated in MSCP. A total of 264 useable reports were collected. Reporting was higher when practices first joined and then decreased. There was an average of 1.4 reports per month. Physicians submitted the majority of reports. Physicians and nurses were more likely to report an incident than office staff. The vast majority of reported incidents were judged to have ‘virtually certain evidence of preventability’ (93%). Harm was associated with 50% of incidents. Only 1% of the incidents had a severe impact. The top four types of incidents reported were documentation (41.4%), medication (29.7%), clinical administration (18.7%) and clinical process (17.5%).
Conclusion MSCP has developed and implemented the first safety learning system in Canada for family practice. All clinic members were encouraged to submit reports, but most of the incidents were reported by physicians. The vast majority of incidents reported were preventable with limited severity. The most frequently reported types of incidents fell into the categories of documentation and medication. The low reporting rates suggest that for family practices incident reporting may not be the most effective method to determine the types and frequency of incidents in family medicine.
- Patient safety
- primary care
- incident reporting
- general practice
Statistics from Altmetric.com
A patient safety incident is an event or circumstance which could have resulted, or did result, in unnecessary harm to a patient.1
A safety learning system (SLS) is a method of monitoring the occurrence of incidents and developing improvement strategies to address the cause of the incidents. It moves beyond a reporting system and towards an environment of continuous learning. In the hospital sector, SLSs have been implemented in many countries2 with later elements being added for family practice. There are multiple reasons why family practice requires a SLS tailored to its unique needs. Family practice has differing organisational structures,3 4 lower acuity of care,3 5 more logistically complex care,6 more prominent patient and social factors,3 6 neither compulsory nor widespread reporting of errors,3 different types of incidents reported5 and a different language.4
No SLSs have been developed in Canada for family practice. Medical Safety in Community Practice (MSCP) is a research program in Calgary, (Alberta, Canada) that developed a SLS specifically for family practice.4 Our system is the first in Canada. In other jurisdictions family practice systems have incorporated only reporting;7–12 ours is one of only a handful that have incorporated both the reporting and learning aspects of an SLS.13–16 Of those that have incorporated both components, none have encouraged all members of the clinic to report; in Hoffman et al only physicians reported15 and in the Wilf-Miron paper reports were from the medical director, malpractice or patient complaints, and were not family practice specific.13
Although learning systems have been virtually ignored in family practice, the reporting aspect has been well studied and multiple classification systems have been developed to describe patient safety incidents.5 17–27 Among all of these classification systems there is a lack of standardisation. Because of the lack of a standard classification system, the WHO developed an International Patient Safety Event Classification (ICPS).28 The conceptual framework for the classification system includes the following primary classes: contributing factors/hazards, incident type, incident characteristics, patient characteristics, detection, mitigating factors, patient outcomes, organisational outcomes, ameliorating actions. There has been no reported use of the ICPS28 for incidents in family practice.
One of the first steps in quality improvement is identifying areas for improvement. Voluntary incident reporting is one of the methods that has been used for identifying areas for improvement.29 30 This method is more successful in some settings than others, with the number of incidents submitted varying from less than one report per clinician or staff member per year11 to eight per doctor per year18 to a high of 45 per doctor per year.12
Until 2006, incident reporting was the most commonly used method to identify incidents.31 Since then, other methods have been successfully used, including trigger tools,32 33 patient surveys,34 35 failure modes and effects analysis36 and chart review.35
The purpose of this paper is to discuss the results of the reporting system of the SLS in family practice. Specifically we report on the number and type of incidents. The learning aspects of the SLS will be discussed in a future paper.
Family practices were recruited via phone calls, business meetings with rural and urban physician leadership, continuing medical education sessions, and faxed and mailed invitations to 958 physicians on the register of the College of Physicians and Surgeons of Alberta practicing in the Alberta Health Services—Calgary zone.
An incident reporting form was developed from an extensive literature review.31 Both content and face validity were determined by piloting the reporting form with physicians, nurses, staff, quality improvement specialists, lay people and co-investigators. The reporting form contained open-ended and multiple choice questions (see online appendix S1). Physicians, nurses, office staff and managers of family practices who volunteered to participate confidentially reported incidents via fax or online to MSCP using the developed form from September 2007 to August 2010.4 Clinics received promotional materials (pens and notepads) and compensation of $20 per report submitted. Bi-monthly meetings were held with each clinic to remind them to submit reports and to discuss and develop improvements to decrease the frequency of incidents.
A more complete description of the development of the reporting form can be found in a previous report by our group.4 Questions on the reporting form, such as type of incident, patient outcome, severity of incident and duration of harm questions, were taken from the WHO ICPS.28 In addition, the provider familiarity with patient scale was a modified version of the scale developed by the American Academy of Pediatrics.37
All reported incidents were reviewed by the MSCP Program Panel, which consisted of researchers, physicians, a quality improvement specialist, a layperson and a nurse. The Program Panel assessed the reports and came to consensus on duplication, incident preventability, patient outcome, patient impact, incidents involving multiple patients, and sequences or cascades. A report was considered a duplicate when the same incident was reported by two or more reporters. Incident preventability was determined using the Preventability Chart reported by Baker et al.38 When more than one patient was involved in an incident, the Program Panel determined whether to split the incident into two or more events or to choose one patient as the primary patient (the one for whom the incident was most significant). A sequence was defined as a continuous or connected series of incidents. A cascade was defined as a type of sequence in which each stage derived from or acted upon the preceding event.39 The researchers evaluated the reports for typology (type of incident).
A total of 19 family practices (15 urban and four rural) consisting of 47 physicians, 53 office staff, 18 nurses, and six clinic managers participated in MSCP. On average, each clinic had 2.9 physicians, 2.7 office staff, 1 nurse, and 0.4 clinic managers. Sixty-three per cent of the participating clinics were computerised.
Table 1 outlines the number and percentage of participants reporting by provider type and the number and percentage of reports submitted. Using a generalised linear model with office staff as the reference it was determined that nurses are 2.1 times as likely to report as office staff (RR 2.08; 95% CI 1.25 to 3.46; p=0.005) and physicians are 1.7 times as likely as office staff to report (RR 1.72; 95% CI 1.08 to 2.76; p=0.023). Clinic managers are not more likely to report than office staff (RR 1.04; 95% CI 0.31 to 3.44; p=0.950). The vast majority of those reporting an incident attended at least one of the bi-monthly meetings (98%).
From September 2007 to August 2010, a total of 270 incident reports were received, six were deemed not to be an incident by the Program Panel, leaving 264 useable reports: six reports included multiple patients, two reports were duplicates, one was a sequence and six formed a cascade. The majority of reports (76%) were faxed, while 24% were submitted online. Most reports were incidents involving female patients (73%). The median patient age was 51 (min 1–max 93) years and the mother tongue of the patient was usually English (94%). The vast majority of reported incidents were judged to be ‘virtually certain evidence of preventability’ (93%).38 Table 2 depicts the reported patient outcomes.
Most incidents were reported to have no severity of impact on the patient (57%), while 24% had a mild impact and 9% had moderate or severe impact. Only 1% of the reported incidents had a permanent duration of impact while 21% had a temporary impact.
Table 3 shows the ICPS type of incidents and the percentage of each type of incident occurring in reports. (Please note, incident reports could be classified into more than one typology.)
On average the clinics participated in the program for 14.0 months (min 1–max 26). They submitted an average of 14.2 reports (min 0–max 40) with an average of 1.4 (min 0–max 7) per month. During this time, the clinics designed, tested and implemented an average of 2.4 (min 0–max 8) improvements.
The number of reports submitted by a clinic decreased over the length of time the clinic was participating in the program. Only one clinic did not follow this pattern. During the first quarter of participation each clinic submitted an average of 5.7 reports, in the second quarter 3.2, third quarter 1.3, fourth quarter 1.2.
In this study there was a wide range in the number of reports submitted by each clinic. The average was 14.2 (0–40). This range is large and it is possible that those clinics reporting more may have had a more open culture that allowed reporting. Clinics also submitted more frequent reports when first starting the program: 5.7 in the first quarter, 1.3 in the third. It is possible that clinics started the study with pressing issues; once these issues were addressed, their impetus for reporting decreased. It is unlikely that reporting decreased due to fewer incidents because clinics relayed unreported incidents to facilitators at the bi-monthly meetings.
The frequency of reporting incidents was low, supporting the conclusion that voluntary reporting may not be the best method of identifying patient safety incidents in family medicine. This is in agreement with Tam et al who reported that the most effective way to identify incidents is through chart review or patient survey and not through voluntary incident reporting.35 Voluntary reporting may not work for a number of reasons, including a perceived lack of time to report, not recognising incidents, not appreciating the value of reporting, and lacking a culture of openness that would promote reporting. Voluntary reporting is expensive. The overall costs of the program over 4 years were $264 000 (paper submitted), giving an average cost per report of just under $1000.00.
Only 38% of family practices in Alberta are computerised and 22% across Canada.40 This lack of computerisation makes it difficult to use other methods of incident identification such as chart audits and trigger tools. There is also a dearth of information on the types of incidents that occur in family medicine in Canada.41 Although other countries such as the USA and Australia9 10 42 have examined their systems in more detail, it cannot be assumed that the incidents found in Canada will be exactly the same as those in other countries. Examining only those areas that have been identified in other countries as high frequency or high risk may miss important types of incidents in the Canadian context.
Our clinic population was similar to the general clinic population within the Calgary zone in terms of size and staff mix. In the general clinic population there are on average 3.5 physicians, 2.5 office staff, 2.3 nurses, and 0.5 managers,40 whereas in the study population there was an average of 2.9 physicians, 2.7 office staff, 1 nurse and 0.4 managers. Fifty-five per cent of study physicians were women compared with 50% in the general population.40 Our population did differ in the use of computerisation; 63% of participating clinics were exclusively using electronic medical records whereas the average in the Calgary zone is 38% and in Canada 22%.40 It is likely that participating clinics were ‘early adopters’, more open to changing their systems than non-participating clinics. This may also account for our difficulty in recruiting clinics. Recruiting for this project was difficult; it took multiple different efforts to recruit sufficient clinics. A total of 948 physicians were invited to participate. We speculate that quality improvement is a new concept for family physicians and that those who participated are on the leading edge of this uptake. Thus it is possible that reporting would be even less in non-participating clinics.
Unlike hospital where most incidents are reported by non-physicians,43 44 we and others45 who have used systems developed specifically for family practice have found that physicians do the majority of reporting in this setting. Further analysis of our data showed that both nurses and physicians were more likely to report than office staff. One of the reasons for physicians reporting more is that the ratio of nurses to physicians is much smaller in family practice than in hospital. The physicians may have reported more often because recruitment was through them, meaning that they had made a commitment to participate. In the hospital, reporting systems tend to be mandated and physicians are less likely to be engaged in the process. Office staff may have been less likely to report as they are hired by the physician as opposed to hospital where staff are hired by the organisation. This difference may cause office staff to be afraid of retribution and thus less likely to report. A higher percentage of office staff (26%) reported in our study compared with Harris et al (7%).45
Although harm was observed in about half of the reports, most of these harms were mild. Similar to the findings of West et al, approximately 9% of patients suffered from severe or moderate harm.46 Unlike Hoffmann et al,15 who reported that 12.8% of incidents led to permanent harm or death, no deaths were reported in our study. In Hoffmann et al only family physicians reported and the authors suggested that the reason for a higher level of the most severe incidents could be due to the misconception in their learning system that incidents were not worth reporting unless something severe happened.15 The lack of death reports to our system may be that either there were no deaths during this time period or incidents leading to death in hospital were not recognised as stemming from an incident in clinic.
In the literature a wide range of preventability for incidents in family physician clinics (23.6–83%) has been reported. Using a preventability scale that was used in the Canadian Adverse Event Study38 we found 93% of incidents had a ‘virtually certain evidence of preventability’. This high level of preventability could be attributed to our very broad definition of an incident as well as the perspective taken while classifying incidents. It could also be due to under-reporting or selective reporting of incidents because we did not attempt to externally validate incidents reported.
Categories and types of errors make it possible to pinpoint potential sites for intervention.47 Although incident reports have the potential to provide a gross approximation of the volume and type of incidents in healthcare settings, their usefulness is limited to providing a glimpse of the scope of the problem.48 We believe that family medicine is not at a stage where incident reports can be used as a means of approximating the volume and type of incidents. During conversations with the clinics, incidents would be discussed that the clinic had not recognised as something that could be reported or was worth reporting; for example, not having the correct contact information for a patient. These unreported incidents were seen more as a ‘fire’ that needed to be put out than a problem that could be reduced through a change in the clinic system. The clinics were not used to looking at their practice from a systems perspective.
This is the first time that the ICPS has been used to classify incidents in family practice.28 The top four categories for incident types were documentation (39%), medication (29%), clinical administration (18%) and clinical process (16%). It is difficult to compare these results with previous studies because the categories are different. In the literature office administration incidents are found to occur between 6% and 51% of the time.7 12 17 41 49 These incidents may be equivalent to our clinical administration and documentation incidents combined (57%). Medication incidents have been reported to occur 14–55% of the time.9 11 12 14 17 26 42 49 Our study results are similar to this finding (29%) and the subcategories of wrong drug, wrong dose, wrong patient are also similar. Diagnostic incidents are reported to occur 6.5–47% of the time.9 11 12 41 42 49–52 Diagnostic incidents are found within the ICPS category of clinical process. Our results for clinical process are in this range (16%). A category of incident often reported by others, but not found within the ICPS incident type category, is communication. Although communication is not an incident type in ICPS it is seen as a contributing factor.28 Contributing factors from our study will be discussed in a future paper.
MSCP has developed and implemented the first SLS in Canada for family practice. Although all clinic members were able and encouraged to submit reports, most of the incidents were reported by physicians. The vast majority of incidents reported were preventable, with limited severity and duration of impact. The most frequently reported types of incidents fell into the categories of documentation and medication. These incident types were very similar to those found in other countries. A reporting system does not appear to be the most effective method to identify areas for practice improvement in family medicine. Future research should concentrate on other methods for identifying areas for practice improvement.
Thank you to our Program Panel members, Dr Leslie Cunning, Dr Stephen Morys, Yvette Penman, Donna Gravistin and Anne Findlay.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
Funding This project is funded by Alberta Heritage Foundation for Medical Research, Canadian Health Services Research Foundation and Canadian Patient Safety Institute.
Competing interests None.
Ethics approval Ethics approval was provided by University of Calgary.
Provenance and peer review Not commissioned; externally peer reviewed.
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.