Article Text
PDF
Abstract
Objectives To report the implementation of a trigger review method (TRM) in primary care, with a particular focus on its impact on patient safety-related findings.
Design Cross-sectional structured review of random samples (n=25) of electronic records of ‘high-risk’ patient groups conducted twice per year (each for a retrospective review period of 3 months).
Setting 274 general practices in two regions of Scotland.
Intervention Contractual incentivisation of TRM implementation.
Main outcome measures Practice participation rate; characteristics of patient safety incidents (PSIs), for example, their prevalence, type, perceived severity and preventability; and actions or intended actions undertaken during and after trigger reviews.
Results 274 of 318 eligible practices (86.2%) returned 536 TRM Summary Reports, which outlined findings from reviews of 13 351 electronic patient records. 1887 (14.1%) PSIs were recorded, with a mean of 3.5 (536/1887) per Summary Report (SD±1.6). Of these, 830 (44.0%) were judged to have caused mild to moderate harm, with 262 (13.9%) cases resulting in more severe harm. A total of 852 PSIs (46.2%) were rated as preventable or potentially preventable. In 459 Summary Reports (85.6%), reviewers indicated implementing one or more improvement actions during the actual TRM process; and 2177 actions after completion of the TRM process (mean 4.1 (SD±3.3) actions per review).
Conclusions The great majority of clinician reviewers ‘successfully’ applied the TRM, uncovering important but previously undetected PSIs, which prompted care teams to take action during and after the trigger reviews. The method and data generated have the potential to drive improvements in related care processes at the practice, regional and national health system level. TRM arguably increased ‘ownership’ of the safety challenge and clinician engagement in implementing their solutions to specific problems identified. Our results suggest that the TRM has potential as a feasible, pragmatic approach to improving primary care safety and quality.
- General practice
- Medical error, measurement/epidemiology
- Patient safety
- Primary care
- Quality improvement
Statistics from Altmetric.com
- General practice
- Medical error, measurement/epidemiology
- Patient safety
- Primary care
- Quality improvement
Introduction
Patient safety is a priority for all modern healthcare systems.1 ,2 However, it is widely accepted that errors do occur, can be preventable and that a significant minority result in patient harm of a temporary nature or more severe in terms of ongoing physical or mental impairment and even death.3–5 Over a decade ago, a seminal UK policy publication recommended that healthcare organisations should systematically learn lessons from patient safety incidents (PSIs), and that system-wide safety interventions were a necessary part of this collective learning and improvement process.2 More recently, Vincent and colleagues stated the case for the importance of identifying and measuring harm as a core patient safety improvement goal,6 while the Francis Inquiry report7 into failings in care at the Mid Staffordshire NHS Foundation Trust in England made multiple recommendations on this issue, including the need for it to be a core organisational priority. The subsequent Berwick Report noted,8 however, that there is currently limited capacity to analyse, monitor or learn from safety-related information at the healthcare organisational level. To date, much of the patient safety agenda has focused on the hospital sector; patient safety is, however, also an issue for general practice/family practice.
In UK general practice, as in other settings, a feasible method of contributing to these learning, improvement and (potential) measurement objectives on a systematic basis is lacking.9 Contributory factors include the limited engagement by general practitioners (GPs) in formal incident reporting systems,9–12 while learning from routine significant event analyses is rarely captured and acted upon either at the level of practices or regional healthcare organisations.13
PSIs in UK general practice—defined here as ‘any unintended or unexpected incident which could have or did lead to harm for one or more patients receiving NHS care’14 are typically reported by patients, identified directly by clinicians or highlighted by colleagues as part of routine practice. However, some safety incident types and latent hazards are not detected so easily, and there is evidence that the majority of incidents remain undetected.15 ,16 Even when PSIs are detected, they are typically not shared within the practice team, nor formally reported and rarely lead to further action.
The recent adaptation, re-design and testing of a trigger review method (TRM) for general practice—also known as a ‘trigger tool’—offer one approach to identifying such events,17 providing practices with an opportunity to reflect and act in the light of events and thus strongly emphasising their importance as change catalysts. Using the TRM to systematically review the clinical records of their own patients for incidents and latent threats provides care teams with a personal perspective of the safety of their care and potentially provides valuable opportunities that are relevant to their practices to take pre-emptive action before harm can occur.18–20
The TRM was adapted from the Institute of Healthcare Improvement (IHI) Global Trigger Tool21 ,22 in 2007, but has since evolved differently in the Scottish general practice context from its originally intended harm measurement purpose to a method of identifying PSIs and, more importantly, improvement opportunities.19 It enables trained clinician reviewers (eg, GPs or practice nurses) to screen and review small, random samples of their electronic medical records of predefined ‘high-risk’ patient groups (eg, those over 75 years with comorbidities and polypharmacy) for PSIs in a rapid, structured and focused manner. This is carried out by searching each of the five common sections of every medical record sequentially for the presence of predefined ‘triggers’, which are clinical prompts or ‘signs’ in the record that may indicate the occurrence of PSIs (box 1). A maximum time of 20 min is allocated per record and three calendar months are retrospectively reviewed in each record. The expectation is that the identification of PSIs will lead to immediate improvements in system processes in some cases (undertaken during the review itself) or serve as a mechanism to guide further action after the review by other means, for example, significant event analysis or criterion audit.13 ,23
A selection of detected patient safety incidents (PSIs) recorded in Summary Sheets
Breathlessness requiring admission in a woman who had her diuretics and ACE inhibitors stopped (previously) during an episode of diarrhoea.
Prescription marked as ‘1 tab daily’—100 tablets issued on each occasion. Incorrectly marked dose, as patient taking two tablets daily for 6 days and three tablets on day 7.
Fall at home—long lie—resulting in confusion, dehydration and acute kidney injury requiring admission. Patient was not taking medication appropriately and using old drugs lying in house.
Eighty-two-year-old female patient was issued with a prescription for atorvastatin 80 mg instead of 40 mg, despite a documented adverse reaction on the 80 mg dose.
Ibuprofen given as acute to already anaemic patient (haemoglobin <10.0).
Delayed diagnosis of incarcerated/strangulated hernia led to prolonged 11-day hospital admission. Earlier diagnosis may have resulted in shorter hospital stay.
Dose of amitriptyline increased to 25 mg (and) original dose not deleted.
Patient with haemoglobin =9.2(and heart failure) untreated and not retested for this.
Patient presented as worried she may be pregnant and taking methotrexate. Pregnancy test negative. No record in the notes that patient had received counselling.
Patient fell due to dizziness. Symptom may have been related to increase in oxycontin for back pain.
In terms of implementation in the UK, evidence of TRM participation can now be submitted as a quality improvement activity as part of GP Appraisal in Scotland,24 while the Royal College of General Practitioners has also included the method as a potential evidence source for revalidation throughout the UK.25 Similarly, it may also inform quality improvement obligations as part of specialty training for general practice.20 In April 2013, the TRM was included in the Scottish version of the pay-for-performance Quality and Outcomes Framework contract; participation in this also doubled as a core contribution to the goals of the Scottish Patient Safety Programme in Primary Care.26 The contract financially incentivised all Scottish general practices (c1000) to apply the TRM on two occasions over a 12-month period and report their findings and improvement actions to their regional National Health Service (NHS) Board.
Despite the rapid adoption and promotion of the TRM, there is little empirical data about the characteristics of the PSIs being identified, and whether its application leads to improvement in the safety and quality of care. This study provides the first known opportunity to examine what is being uncovered and achieved across regional care systems by reporting the findings from two Scottish NHS Board areas that implemented the TRM in general practice. Our aims are therefore twofold:
To report the characteristics of the PSIs that were detected and reported, including their prevalence, type of care processes they relate to and their perceived severity and preventability.
To report the actions and intended actions undertaken during and after the trigger reviews.
Methods
Design
A cross-sectional review of 25 electronic patient records (for a retrospective period of 3 months) conducted twice per year by clinicians trained in using the TRM.
Setting and sample
The study was set in two Scottish NHS Board areas containing 56 and 262 general medical practices, respectively, and providing health and social care to around 30% of the Scottish patient population. Both regions include a range of socioeconomic, semi-rural, urban, suburban and inner city settings. Participating practices were those undertaking the TRM as part of contractual incentivisation and reporting these data to their NHS Board.
TRM training
One or more clinicians nominated by the practice (typically a GP and practice nurse) were trained in the TRM by an informed GP while attending local collaborative learning events arranged by each NHS Board. Training typically lasted 1.5 to 3 h and involved an audio-visual presentation and a practical TRM exercise using simulated medical records containing ‘hidden’ PSIs for individual delegates to detect. This was followed by feedback and open discussion. A range of educational material, including a practical application guide and detailed examples of previously completed reviews, were also provided for reference.
Conducting the TRM
Two trigger reviews were conducted, at least 6 months apart, during April 2013 to March 2014 by a clinician trained in the TRM. Previous TRM pilot testing has demonstrated the validity of the triggers and the ability of clinical reviewers to detect PSIs using this approach.17 ,19 ,20 ,23 The process was implemented as follows:
The practice chose a ‘high-risk’ patient group on which to conduct the TRM review from a list of examples provided and previously published.23
A practice team member (eg, practice manager, administrator or clinician) compiled a list of all relevant patients from the practice's clinical information system.
A random sample of 25 records was identified from the list (using a random number generator or similar process).
A clinician trained in the TRM process then reviewed a 3-month period in each electronic patient record. The review time period was any three consecutive calendar months in each patient record before the actual review dates.
Data were collected using a predesigned pro forma (see below), which also contained details of each of the clinical triggers as a cognitive aid to guide a rapid review process.
Practices repeated this process when conducting the second review 6 months later, although they had the option to decide whether to review another random sample of the same patient group, or alternatively choose another ‘high-risk’ group.
Data collection
Data were recorded by reviewers on a standardised Summary Report pro forma (see online supplementary appendix 1) for each batch of 25 medical records reviewed. The pro forma is structured over two pages to guide reviewers through the three consecutive steps of the TRM: (1) planning and preparation; (2) reviewing the records; and (3) action, reflection and learning. Data collected include the number of detected ‘triggers’, details of PSIs uncovered (including judgements on perceived severity and preventability), learning needs identified and immediate or future actions that were or should be taken to minimise risks of PSI re-occurrence, and length of time taken to conduct the review. Practices were advised that if five PSIs were detected before the full 25 records were reviewed then this was considered a sufficient workload and the remaining time of the review should be allocated to improvement actions. Practices submitted their completed Summary Sheets to the local NHS Boards as evidence of TRM completion (for financial reimbursement) and to enable data analysis to facilitate local learning.
Supplementary appendix
Data analysis
TRM data were provided to the authors for analysis by both NHS Boards. Data from all the submitted pro forma were extracted to a Microsoft Excel spreadsheet by a single administrator, and the PSIs were then coded (by CdW, CB and SL) and analysed (by CdW) using simple descriptive statistical methods. The types of reported PSIs were categorised jointly by the authors through an iterative process of descriptive coding of the recorded incidents and consensus building to resolve disagreements.
Quantitative data were analysed using simple descriptive statistics (eg, frequency counts and percentages). Free text entries were grouped into major themes.
Results
Response rate
A total of 274 of 318 eligible practices (86.2%) from the two NHS Boards returned 536 Summary Reports, which outlined findings from reviews of 13 351 electronic patient records. In NHS Board ‘A’, 45/56 practices returned at least one completed Summary Report (80.4%), with 11/56 failing to return any (19.6%). Of these submissions, 74 reports from 44 practices (97.8%) were suitable for coding and these summarise the findings from 1843 electronic patient records. For NHS Board ‘B’, 229/262 practices (87.4%) returned a total of 462 Summary Sheets that were suitable for coding, while 7/262 (2.7%) returns were unsuitable for coding and 26/262 (10.0%) did not return any. The 462 reviews summarise the findings from 11 508 electronic patient records, with the vast majority of reviews (92.7%) being undertaken by GPs.
Patient safety incidents
Detected PSIs
A total of 1887 PSIs were reported, which equates to 14.1% of all records reviewed and a mean of 3.5 (536/1887) per Summary Sheet (SD±1.6). NHS Board ‘A’ reviewers documented 251 PSIs (13.6%) and a mean of 3.4 PSIs per Summary Sheet (SD±1.6). In 3/74 Summary Sheets (4.5%), reviewers did not record a single PSI. For NHS Board ‘B’, reviewers reported a total of 1636 PSIs (14.2%), with a mean of 3.5 (SD±1.6) PSIs per Summary Sheet. In 21/262 Summary Sheets (4.5%), reviewers did not record a single PSI. Selected examples of detected PSIs are shown in box 1.
Characteristics of PSIs: severity, preventability and type
For both NHS Boards combined, a total of 830 PSIs (44.8%) were judged to have caused mild to moderate harm to patients, with 262 cases (14.2%) resulting in more severe harm. A total of 852 PSIs (46.9%) were rated as preventable or potentially preventable (table 1). Details of the most common types of PSIs are outlined in table 2. The 10 medications that were most frequently implicated are included for interest.
The severity and preventability of patient safety incidents as rated by clinical reviewers in both National Health Service Boards
Frequency and details of main types of patient safety incidents (PSIs) and the most common medications implicated
Improvement actions
Actions undertaken during reviews
Reviewers indicated undertaking one or more actions relating to the safety and quality of care during trigger reviews in 459/536 trigger summaries (85.6%). There were four main types of immediate actions, which relates to the following care processes in general practice: (i) coding and record keeping; (ii) prescribing and medication; (iii) communication; and (iv) further investigation and follow-up. Examples of the types of actions and a selection of verbatim illustrative quotes from the summary sheets are shown in table 3.
Types of actions undertaken during trigger reviews with selected examples and quotes
Actions and intended actions undertaken after trigger reviews
Overall, reviewers undertook or intended to undertake 2177 actions subsequent to and as a result of performing 536 trigger reviews, with a mean of 4.1 actions after each trigger review (SD±3.3). The most common types of actions were ‘feedback to colleagues’ (41.2%) and ‘add to appraisal’ (18.1%) while the least common was using the Plan-Do-Study-Act (PDSA) method (1.5%). The number and types of specific actions are shown in table 4.
Summary of the action(s)* reviewers indicated that they undertook or planned to undertake after the trigger reviews
Reported trigger review times
The overall time to conduct reviews (of a set of 25 records), implement or consider action and complete Summary Sheets for both NHS Boards was 86 253 min (n=495 reviews) with a mean of 174.2 min/trigger review (SD±78.4 min, range 30 to 480 min).
Discussion
This study reports the main outcomes from the implementation of the TRM in general practices across two NHS Health Boards in the west of Scotland. The vast majority of general practices participated in this patient safety improvement initiative and applied the TRM to the medical records of their own patients. The main findings are that most clinician reviewers were able to detect PSIs in the medical records of their patients and were willing to report these. In addition, the majority reportedly took or intended to take a wide range of formal and informal improvement actions, including protocol revisions, significant event analyses, clinical audits, sharing learning points in their teams and editing and updating medical records to minimise risks.
The majority of detected incidents were categorised as low-to-moderate severity, ‘near misses’ and latent system hazards. The types of PSIs uncovered by the TRM will be familiar to most patient safety researchers and indeed primary care clinicians. We would argue, however, that the main value of the reported PSIs was derived from them being detected within the records of the reviewers’ practices in a ‘real-world’ setting—rather than their type or frequency. The TRM process potentially facilitated clinician engagement by providing opportunities to tackle issues that are perceived as relevant to individual practice teams and clinician reviewers. This facilitation of potential clinician engagement is arguably one of the crucial differences between the primary care TRM and other similar improvement interventions and may promote its sustainability and routine use in primary care settings where adequate incentivisation and protected time are available.
Typically, frontline staff first need to assure themselves that there is a real and important problem that relates specifically to them. In NHS Scotland, GPs are independent contractors with the freedom to choose whether they participate in safety improvement initiatives or not. In our experience the vast majority perceived patient safety as important, but only one of a number of priorities. GPs' participation in an initiative increases if they understand and are able to see quick, unambiguous and practical evidence that an intervention has value for them and their patient population.
If we assume that the reported study data is indicative of the TRM ‘working’ in these two NHS Boards, then it would be fair to say that with its current national implementation it should detect PSIs in all other NHS Board regions. Moreover, based on the scale of our findings, there is a possibility of this approach having demonstrated that it is potentially the most effective of all current methods for detecting, reporting and acting upon patient safety-related issues in the Scottish (and by extension UK) general practice setting, particularly since there is very low engagement in formal incident reporting systems.9 ,13 With the current focus on patient safety,26 we suggest that the TRM is a potentially feasible safety improvement intervention, which could be applied more widely in UK general practice and international family practice settings, either as part of a large-scale safety improvement initiative or voluntarily as a simple quality improvement activity.
However, implementing and sustaining safety interventions in complex healthcare environments is problematic for a whole raft of cultural, social and resource-based reasons.27 It is suggested that intervention success is predicated on three common conditions all being met: (1) making it straightforward to do; (2) measuring and providing feedback on related outcomes; and (3) normalising the intervention as part of routine work to improve performance and change culture.27 ,28 Our findings provide some evidence of the ability of care teams to understand and apply the TRM to detect PSIs at the practice level (Condition 1) and the potential for organisational level monitoring and measurement to feed back learning from safety-related issues across practices (Condition 2). However, for the full benefits of the TRM to be realised it would require to be routinely implemented, perhaps in a very targeted way. One example might be to measure and reduce avoidable harm from Warfarin toxicity. However, this may require linkage with multiple intervention strategies (such as comparative and highly visible audit and feedback) to drive improvements on a regional and national basis (Condition 3).
As part of recent related research on the TRM potential to reliably measure harm rates in general practice,29 a basic formula was designed using computer-generated simulation to inform sample sizes (the minimum numbers required are similar to those reported in this study) and acceptable levels of statistical precision and power necessary for the purpose of detecting whether improvements in safety performance are being realised. Therefore, by repeating the implementation approach reported in this study on a targeted patient subpopulation there is the potential to test ‘Condition 3’ by measuring, monitoring and driving improvement on a regional and national basis. Vincent et al's6 recently published framework for the measurement and monitoring of safety supports this approach by calling for more ‘specific and nuanced’ measures of harm of direct relevance to different care settings to be developed.
Practical implications
Besides potentially providing key insights into safety and risk matters, application of the TRM importantly may also serve to facilitate practice and organisational learning through integration with quality improvement methods such as significant event analysis,13 plan-do-study cycles,30 clinical care bundles31 and criterion-based audit.32 For example, to assess in much greater detail system-wide human factors interaction issues contributing to delayed diagnoses, suboptimal therapeutic management, disease and drug monitoring, and problems with computerised support technologies. This is particularly helpful given that appraisal and revalidation requires GPs to analyse two significant events per year (with the UK General Medical Council encouraging these events to be PSIs rather than broader quality-of-care issues). Identification and analysis of these previously undetected PSIs are particularly pertinent to improving the opportunity cost of significant event analysis topics.
Strengths and limitations
There was high-level engagement from practices in both NHS Board areas in response to incentivisation to apply the TRM, which provided significant levels of data on captured PSIs. Based on this study, the TRM can clearly identify PSIs, which provide evidence of its validity and appear to have been ‘tolerated’ by large numbers of clinicians trained in the process (evidence for professional acceptability of the method) in a relatively short timescale (evidence for feasibility). However, these data are self-reported and so subject to bias as there was no means to externally verify data quality, particularly in terms of judgements on what constitutes a PSI, how severe it is and if it is preventable. Inevitably there will be inter-rater variations on these issues.
Conclusion
Our understanding of the types and nature of the threats to patient safety is continually increasing.33 Patient safety research in primary care is beginning to evolve and demonstrate the need for co-ordinated action to make systems safer5 ,9 ,13; and from this perspective the rationale for improvement seems obvious. However, much of the primary care workforce is unaware of the evolving evidence base, nor do they necessarily share this perspective. Consequently, efforts to harness the patient safety literature in order to meaningfully improve care have so far had limited demonstrable effect.34 In response, we developed, tested and implemented the TRM and propose this approach as a pragmatic bridge between (i) research/academic general practice and clinical service delivery; (ii) national and individualised approaches to improve patient safety; and (iii) initiatives focusing only on measurement or improvement. There may not be any patient safety panaceas yet, but our findings suggest that the TRM is a potentially feasible approach to identifying and addressing a range of important safety issues that were previously undetected in the medical records of ‘high-risk’ patient groups in general practice.
References
Footnotes
Correction notice This article has been corrected since it was published Online First. The name of the fifth author has been corrected.
Twitter Follow Paul Bowie at @pbnes and Catherine A O'Donnell at @odo_kate
Contributors CdW: concept, study design, data analysis, co-development and critical review of manuscript. CB and SL: data analysis and critical review of the manuscript; JM and CAO: contribution to study design and critical review of the manuscript. PB: concept, study design, co-development and critical review of manuscript, study guarantor.
Funding NHS Education for Scotland.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
TRM guidance and information Education guidance and more information on TRM can be accessed online here: http://www.healthcareimprovementscotland.org/our_work/patient_safety/spsp_primary_care_resources/trigger_tool.aspx