Physical separation

Complaints handling is traditionally located near hospital lawyers, patient advocates or guest services, rather than the later-developed quality and safety departments. It may also be difficult to reference complaints data for QI purposes as complaints are often archived in binders, sorted on patient or physician names, rather than in accessible digital databases. This separation of complaints from QI practices precludes this information from being used, for example, to gain insights into patient-centredness or continuity of care.6

Case-by-case handling

Moreover, while most hospitals have installed systems to learn from adverse events and incidents, systems to learn from complaints are lacking.7 The ability to learn from complaints is particularly limited by the one-by-one approach to complaints. On receiving a complaint, most hospitals notify the involved providers, who (help to) write a response.8–10 While important for restoring the provider–patient relationship, this approach also treats complaints as isolated issues between individual providers and patients. One negative effect of this is that providers could be given the feeling that they are individually responsible for the whole of the negative patient experience, inducing feelings of shame and guilt that hamper learning.11 Another result is that the case seems ‘closed’ once a response is sent, which fails to trigger a deeper investigation, and to share the learning within the team, department or hospital. The handling process may be finished when the complaint and response letter are archived, but this does not necessarily mean that the complaint’s problem has been resolved. It will remain difficult to define or determine when exactly a complaint is resolved, but complaints handling should be seen as the beginning of a process to gain a deeper understanding of the patient’s concerns and how the issues could be addressed, rather than the end of a service provided to patients.

Elusive source of information

As complaints often result from cascades of problems until ‘the straw that broke the camel’s back’,12 there will often be various issues and sources of frustration that contributed to the patient’s negative experience. Identifying the exact underlying problems can be remarkably difficult. Complaint letters can be difficult to read as they are mostly unstandardised, unstructured and emotive,13 as well as written by patients and families with varying educational backgrounds. Moreover, complainants may have interpreted certain events more harshly in a context of cumulative hurt and frustration,9 and may focus on subjective aspects of care, such as compassion, while leaving other important contributory factors or problems undiscussed.13 This process is further challenged by bias on the receiving end, as it can be difficult to interpret complaints in a non-judgemental, unbiased manner, particularly when one’s skills and attitudes are criticised.13 A survey among physicians showed that one in three did not consider complainants ‘normal people’, and this was even more so among physicians who had experienced complaints.14 This raises the question whether we should rely on individual providers to draw lessons from complaints. At the same time, close involvement of healthcare professionals is essential to provide medical and context knowledge, as well as for front-line engagement, which are required for learning and improvement processes. Both receptivity to complaints and patient’s willingness to report complaints might benefit from using a term such as ‘patient feedback’ instead of a term that is synonymous with ‘whining’ and ‘moaning’ in the dictionary.

Coding tools do not identify underlying problems

While methods for standardised analysis of complaints seem beneficial to structure and categorise the problems addressed, this process is first of all complicated by the complexity of these letters, as described above. Moreover, coding strategies used by these methods may pose problems when trying to use complaints for QI. A recently validated tool is the Healthcare Complaints Analysis Tool, which was developed using taxonomies from 59 previous studies.6 13 This method requires taking complaint letters at face value, strictly adhering to the words used in the text. It is understandable that the method does not allow personal interpretation of the letter’s content, but this type of coding may therefore point to different issues than in-depth investigations of the situation and context would. For example, if a complainant writes that he received the wrong treatment, this will be coded as ‘clinical’, while in fact the underlying problem may be related to insufficient explanation and hence ‘communication’. While contextual information may be available in the provider’s response letter, these are not taken into account in these analyses.

Healthcare is a team effort

It is not easy to determine whether problems that triggered a complaint are individual-oriented or system-oriented, as relationships between individual-related and system-related causes of problems are complex and difficult to separate.15 16 Some complaints about individual behaviour may be related to underlying system problems. To illustrate, complaints about a brusque doctor or a nurse not responding to call bells in a timely manner could reflect the typical behaviour of that clinician but could also reflect problems with problematic workload or conflicting expectations of staff. A tendency to view the individual provider as the problem that needs fixing fails to identify underlying system factors17 and could also unnecessarily damage healthcare professionals. While it has been shown that a small number of physicians account for a large proportion of complaints,18 we cannot rule out whether these providers were more prone to complaints due to a larger volume of patients or more difficult patients,16 19 such as patients with a greater risk of complications and hence of complaints.20 The use of complaint rates as a metric to identify ‘bad doctors’ would therefore, as most tests, render false positives, falsely accusing colleagues of incompetence. Moreover, it seems unrealistic to regard patient complaints as criticisms of individual providers. Modern healthcare is provided by many hands,21 and providers are part of larger teams and systems. It has been estimated that medical and surgical patients may see up to 44 and 75 different health professionals during their hospitalisation.22 Accordingly, complaint letters frequently address more than one issue,6 related to providers of several disciplines (including administrative staff), departments or institutions. As a result, it often remains unclear what the exact target group for improvement should be.

Estimating the size of the iceberg

Another distinction that should be made is whether the problem addressed in the complaint was an isolated occurrence or a recurring issue, as we would not want to make changes to our systems on the basis of very specific, rare events. Such corrective actions may also harm our ability to perform well, for example by increasing complexity. It is logical that a recurring problem would require a different response than a single ‘mistake’, but the difficulty lies in (determining who is responsible for) making these judgements, as also discussed in the Just Culture theory.23 Unfortunately, we cannot rely on complaint rates to distinguish isolated from recurring problems, as the likelihood that a problem triggers a complaint is not purely related to the frequency of the underlying problem. In other words, a single complaint may represent only ‘the tip of the iceberg’ for a problem, but may just as well represent an unfortunate, rare occurrence. In fact, complaints are hetereogenous5 and relations with underlying problems are complex. One type of complaints may have different underlying problems, requiring different improvement strategies (eg, discharge complaints can be triggered by communication as well as logistic issues). At the same time, one underlying problem could trigger different complaints (eg, understaffing triggering complaints about staff behaviour as well as clinical care quality). Without deeper investigation, most complaints may simply trigger clinical or communication skills training.

Infrequent and imperfect data

As complaints are so infrequent overall, with rates from the literature ranging between 0.1% and 0.9% of all admissions,20 24 25 it might take a while before issues recur, by which time circumstances and opportunities for improvement may have changed already. Analogous to incident rates,26 27 these numbers will be affected by many other factors than quality of care, such as patients’ access to the complaints process. These features make complaint data unfit for monitoring, limiting our ability to identify when a complaint-based improvement is successful. In addition, some patients may be reluctant to file a formal complaint and more inclined to report their concerns in a patient survey instead. Triangulation of complaints data with data about negative patient experiences, for instance extracted from hospital surveys, may establish sufficient volumes and seems an alternative approach worth considering to facilitate learning from the patient’s perspective.

Using complaints as triggers

As caring for patients is a team effort, learning from patients’ complaints should equally be a collaborative process. Such an approach underlines that ‘a complaint against one of us is a complaint against us all’ and encourages sharing the learning with colleagues. This would require transforming complaints handling from a service provided to patients, where the learning remains with the responding clinician, into a joint effort of clinician teams and QI staff. For example, complaints could receive a more prominent role in existing learning practices, such as morbidity and mortality conferences. This would encourage discussing the patient’s perspective as well as soft skills, such as communication or empathy, at these meetings. These meetings could also provide a forum for peer support, and peer feedback, which has been demonstrated to reduce complaint rates of individual providers.28 These discussions may also help to determine whether problems addressed in complaints are recognised as recurring problems. Yet some colleagues may be reluctant to report problems with a fellow clinician.29–31 Therefore, additional investigation will likely be required to assess whether problems raised in complaints are also reflected in other available sources of information, such as interviews with complainants and providers, direct observation of care,17 or review of response letters and medical records. Moreover, hospitals could use triangulation with other data from the clinician or patient perspective, such as quantitative outcome or patient experience data (eg, National Surgical Quality Improvement Program [NSQIP] or Hospital Consumer Assessment of Healthcare Providers and Systems [HCAHPS]), but even then it will remain extremely difficult to identify and solve the actual system problems, a process that requires substantial work and investments.17

To conclude, there are various barriers that hamper using patient complaints to develop improvements. Yet complaints could be taken out of isolation and more closely connected to other QI processes. The associated costs and efforts will vary per hospital, as, for example, not all hospitals have digitalised complaints8 32 33 or routinely collect patient experience data. Hospitals could start by using complaints as triggers for participative learning in teams and further in-depth enquiry with other available QI data. This would address at least part of the various challenges described in this paper and allow to share the learning from patient complaints within teams and institutions.

• Patient complaints represent patients’ perspective on healthcare, but are hardly used for improvement, which is likely influenced by various problems related to this specific source of information.

• Complaints are handled in isolation on a case-by-case level, which fails to trigger deeper learning or investigation and to align with improvement and learning practices.

• Complaint letters are an especially complex and elusive information source with data of low and unreliable volume, which challenges efforts to categorise and code these data, and thereby complicates identifying underlying problems and adequate improvements.

• These features create difficulties to determine whether problems addressed in complaints are individual-related or system-related, and whether these reflect an isolated or recurring issue, which all have implications for QI efforts.

• Given these problems, complaints should be used as a starting point for collaborative learning and used as triggers for further enquiry with other QI data, such as patient experience data.