The American Journal of Managed Care September 2012
Frequency of and Harm Associated With Primary Care Safety Incidents
Objective: To assess frequency and severity of patient safety incidents in primary care.
Study Design: Cross-sectional survey of health- care professionals in Swiss primary care offices.
Methods: Physicians and nurses in primary care offices were surveyed about the frequency and severity of 23 safety incidents. Differences between professional groups and types of offices were analyzed. Reported incidents were classified in a matrix.
Results: A total of 630 individuals (50.2% physi- cians, 49.8% nurses) participated. Among them, 30% of physicians (95% confidence interval [CI] 25%-35%) and 16.6% of nurses (95% CI 12%-21%) reported that at least 1 of the incidents occurred daily or weekly in their offices (c2 16.1, P <.001). On average, each responder reported a total of 92 incidents during the preceding 12 months (mean of 117 events for physicians, mean of 66 events for nurses; P <.001). Documentation failure was reported most frequently.The highest fraction of last occurrences with severe injury or death was for diagnostic errors (4.1%). Unadjusted for caseload, staff working in medical centers reported higher frequencies of several incidents. The frequency-harm matrix suggests that triage by nurse at initial contact, diagnostic errors, medication errors, failure to monitor patients after medical procedures, and test or intervention errors should be prioritized for action.
Conclusions:This study presents a supplemental approach to identification of safety threats in primary care. Many incidents occur regularly and are highly relevant for healthcare professionals’ daily work.The results offer guidance on setting priorities for patient safety in primary care.
(Am J Manag Care. 2012;18(9):e323-e337)
- Because incident reporting varies among professional groups and types of offices, it is essential to involve physicians and nurses and to consider different types of office organization when safety incidents in primary care are studied.
- In this sample of primary care offices, triage by nurse, diagnostic errors, failure to monitor patients after medical procedures, test or intervention errors, medication errors, and documentation failure were identified as “hot spots” that warrant further action.
Incident reporting has a long tradition in clinical risk management and is increasingly used in outpatient care.5-7 Indeed, incident reporting has been the dominant method for study of safety incidents in primary care.8 It is based on voluntary and usually anonymous reports of physicians and nurses and is being used to describe the types and characteristics of patient safety incidents. These reports may vary considerably with respect to the included information, and the likelihood of “true” incidents being reported is unclear. Studies based on this method describe strong variations in the number of reports submitted.9 Moreover, professional groups differ in their frequency of reporting. In contrast to hospital settings, in primary care physicians report considerably more incidents than nurses do.10,11 O’Beirne et al concluded from a very low report rate (<1 report per person per year) that incident reporting may be a costly but not very effective way to study safety problems in primary care.10
With chart review, medical records are analyzed by independent experts in order to identify adverse events and to assess potential harm and preventability in each case.12 This analysis requires complete and correct patient documentation to provide valid results. In many cases relevant information may be unavailable.13 As chart review is a time-consuming approach, many resources are needed to analyze a large number of patient records at different primary care offices. This large sample is necessary to cover a broader range of incidents and contexts.
What chart review and incident reporting have in common is that it is impossible to systematically cover information on the entire range of safety events in medical offices. Instead, these methods cover events with a relative high likelihood of being reported or being documented in the medical records (eg, adverse drug events, errant drug prescriptions). The main aim of this study is to apply a supplemental method to gain additional insight into safety hot spots in primary care. Using survey methodology, we assessed the frequency and severity of incidents in the offices based on the experience of both physicians and nurses. We classified events in a frequency-harm matrix that may be useful for identification and prioritization of safety hot spots.
In a cross-sectional study, primary care physicians and nurses working in outpatient offices were surveyed about the occurrence of safety problems in their offices. A preliminary list of safety incidents was derived from empirical studies.1-3,5-7,9,14-22 We also analyzed incidents reported to Swiss primary care critical incident reporting systems. After discussion with primary care safety experts, physicians, and nurses, the list was concomitantly adapted and finally included 23 potential incidents. The incidents were organized along the care continuum in chapters titled diagnostic process, medication, other therapeutic and preventive measures including interventional procedures, patient encounter and information, organization and work flow in the office, cooperation with other providers, and storage of drugs and materials. Each incident was accompanied by 2 sensitively selected brief examples to guide respondents, explain the range of problems covered by the class of incident, and support memory. Respondents were asked to respond to 2 questions for each potential incident: (1) How frequently did the incident occur in their office during the past 12 months? Response categories were “daily,” “weekly,” “monthly,” “yearly,” “never,” and “never in the past 12 months but at least once before.” Respondents were instructed to report all incidents they were aware of irrespective of their personal role in the incident. (2) What was the severity of harm associated with the last occurrence of the incident in the office (“last occurrence”)? The response codes were “no harm,” “minor harm,” “moderate harm,” “severe harm,” and “death.” Concise definitions of the severity ratings were provided and respondents were instructed to consult these before answering (Appendix A). We also assessed the safety climate in the survey (data not reported). Background information on the respondents and their working environment was obtained as well. An iterative pretest was conducted with 8 practitioners, and the survey was subsequently revised.
The sample consists of all primary care physicians (n = 627) formally organized in 4 large physician networks and included 472 offices. The physician networks were selected because they cover a broad range of office types and regions in the German-speaking part of Switzerland including 2 metropolitan areas, smaller cities, and rural regions. Two networks oblige their members to adhere to guidelines and quality-ofcare standards (eg, participation in quality circles). The others predominantly collaborate on organizational and financial issues. The sampled physicians were mailed the questionnaire by post, together with a cover letter and prepaid envelope. Each physician received 2 complete survey sets. The physician was instructed to pass the set labeled “nurse” to 1 nurse according to the alphabetic position of the first letter of the last names of all nurses working in the office. The study was promoted by representatives of the networks (eg, medical directors, quality managers) before and during data collection. One network offered compensation for participation (approximately US $45), but very few offices (<5) requested it. A few days after the mailing, nurses in each participating office were sent a letter to inform them about the study, announce the survey, and explain how nurses would receive the questionnaire by the physician. The survey was completely anonymous. A reminder including an identical set of questionnaires was sent after 2 weeks to the entire sample. Ethical approval is not necessary for this study in Switzerland.
Frequency and severity of reported incidents were analyzed descriptively. Chi-square tests and t tests were conducted to test for group differences in frequency and severity of incidents between occupational groups (physicians vs nurses) and types of office (single handed vs joint practice vs medical center). All tests were 2-sided and P <.05 was considered significant. We did not adjust for multiple testing.
To support interpretation and usability of results, we developed a 2-dimensional frequency-harm matrix similar to the Veterans Health Administration Safety Assessment Code Matrix, the matrix used by the UK National Patient Safety Agency, and to other classification systems.23,24 Such classifications are used to categorize potential safety incidents and to prioritize hazards for risk management. Rather than using judgments of likelihood and consequences of potential events, we used the quantitative data provided by respondents in our study to plot the reported frequency of incidents against the severity of the last occurrence. As each incident was described by 2 response distributions (frequency and severity),2 analytical steps were conducted to collapse and reorganize the data.
Frequency. First, the frequency of incidents was transformed to the mean number of occurrences of each incident reported per respondent. The number of responses in the daily category were multiplied by 308 (the mean number of office days per year), weekly occurrences were multiplied by 52, monthly occurrences were multiplied by 12, and yearly occurrences were multiplied by 1. Incidents that never occurred were multiplied by 0. The sum of these events was divided by the number of respondents per incident. This average number of incidents per respondent and year was then classified as “singular” (<1 event per year per observer), “rare” (1-4 events per year per observer), “occasional” (4-6 events per year per observer), “probable” (6-12 events per year per observer), and “frequent” (>12 events per year per observer). Because responders reported incidents in their offices, these were inn cidence rates per observer with the office as the space of observation.
Severity. Second, the distributions of the severity ratings of last occurrences were transformed to 4 categories: “catastrophic” (at least 1 death reported), “critical” (at least 1 episode of severe harm reported, but no death), “moderate” (>25% of respondents reported minor or moderate harm, and none reported death or severe harm), and “marginal” (>75% of respondents reported no harm, and none reported death or severe harm).