Re: Prevalence, severity, and nature of preventable patient harm across medical care settings: systematic review and meta-analysis
I would like to thank all authors who submitted responses to this systematic review and meta-analysis.1 Their positive remarks as well as criticism encourage a fruitful debate about the methodological challenges in examining the prevalence of preventable patient harm in medical care settings.
Hamilton makes a number of valid queries reflecting key methodological challenges, which we had to resolve while conducting and reporting this systematic review and meta-analysis.
There are certainly variations in the number of patients treated across primary care, general hospitals and other medical care specialities. However, this is a meta-analysis of international studies with a varying flow of patients across different medical care settings/specialities. Distinguishing the proportion of patients in each medical care setting would be impossible due to these country-level variations. Excluding studies conducted in intensive care and surgical care units would indicate selective reporting.
Instead, the following two strategies were applied in this meta-analysis. First, the pooled prevalence estimates of preventable patient harm in each medical care setting/speciality are reported in Figure 2 to highlight such variations (see Figure 2). Second, because the prevalence of preventable patient harm was expected to vary from very small to large across studies, the Freeman-Tukey Double Arcsine transformation was used to stabilise the variances2 and then perform a random effects meta-analysis implementing the DerSimonian-Laird method.
The pooled estimate of the prevalence of preventable patient harm is primarily driven by data from general hospitals (where the vast majority of studies were based) rather than data from intensive care or the surgical care units. The headline conclusion that 1 in 20 patients experience patient harm stands even if intensive care and surgical care units are excluded from the analyses (although this exclusion is not justified). We also highlight the need to increase the evidence on the prevalence, nature and severity of preventable patient harm in primary care and other busy medical care settings such as psychiatry.
The denominator in this meta-analysis is the number of patients or unique patient consultations. Studies which did not provide data on the number of patients exposed to harms or number of unique consultations containing harms (or we were unable to calculate/obtain this figure) were ineligible for inclusion in this meta-analysis. We have purposively chosen this approach to avoid double counting of patients who may have had more than one consultations and experience more than one preventable harms.
The timing of studies was assessed as a possible covariate in the meta-regression analyses. Large timing variations were found across studies. For example, in some studies the medical records of patients were reviewed for only one week whereas a 12-month review of medical records were undertaken in other studies. The classification of the various timings reported across studies into a manageable number of covariate values was not possible. According to the recommendations of Thompson and Higgins, each covariate value should be based on a minimum of 8-10 studies.3 Consequently, we did not formally include timing as a covariate in the meta-regression analyses but we outlined this important methodological variation across studies as a limitation in the discussion.
Zermansksy highlights the challenges in detecting and assessing the causality of patient harm. Assessing the prevalence of overall patient harm is a scientifically important question and future triangulation of methods could improve accuracy e.g. by providing a better framework or using a combination of methods (e.g. mixed-method approaches with parallel involvement of patients as partners).4-7 However, despite efforts and progress, some patient harms are undetectable and particularly challenging to assess their causality. By definition, preventable harm occurs as a result of an identifiable modifiable cause, and/or its future recurrence can be avoided by reasonable adaptation to a process, or adherence to guidelines. 8 Thus, the fundamental advantage of focusing on preventable patient harm is that improvement strategies can specifically target those types of harms that are identifiable and avoidable.
1. Panagioti M, Khan K, Keers RN, et al. Prevalence, severity, and nature of preventable patient harm across medical care settings: systematic review and meta-analysis. 2019;366:l4185. doi: 10.1136/bmj.l4185 %J BMJ
2. Freeman MF, Tukey JW. Transformations Related to the Angular and the Square Root. Ann Math Stat 1950;21(4):607-11. doi: DOI 10.1214/aoms/1177729756
3. Thompson SG, Higgins JPT. How should meta-regression analyses be undertaken and interpreted? Stat Med 2002;21(11):1559-73. doi: 10.1002/sim.1187
4. Vincent C, Taylor-Adams S, Stanhope N. Framework for analysing risk and safety in clinical medicine. 1998;316(7138):1154-57. doi: 10.1136/bmj.316.7138.1154 %J BMJ
5. Hignett S, Lang A, Pickup L, et al. More holes than cheese. What prevents the delivery of effective, high quality and safe health care in England? Ergonomics 2018;61(1):5-14. doi: 10.1080/00140139.2016.1245446 [published Online First: 2016/10/08]
6. Jylha V, Saranto K, Bates DW. Preventable adverse drug events and their causes and contributing factors: the analysis of register data. International journal for quality in health care : journal of the International Society for Quality in Health Care 2011;23(2):187-97. doi: 10.1093/intqhc/mzq085 [published Online First: 2011/01/19]
7. Sujan MA, Ingram C, McConkey T, et al. Hassle in the dispensary: pilot study of a proactive risk monitoring tool for organisational learning based on narratives and staff perceptions. BMJ quality & safety 2011;20(6):549-56. doi: 10.1136/bmjqs.2010.048348 [published Online First: 2011/03/15]
8. Nabhan M, Elraiyah T, Brown DR, et al. What is preventable harm in healthcare? A systematic review of definitions. Bmc Health Serv Res 2012;12 doi: 10.1186/1472-6963-12-128
Competing interests: No competing interests