Deductive coding: question type

Inductive coding: perceived knowledge gaps

In addition to coding the questions deductively based on ICPC-2 PLUS and the taxonomy of generic clinical questions, we conducted a qualitative content analysis of questions to identify perceived knowledge gaps. Coding was performed inductively with categories derived from the data.14 Inductive codes were collected by senior author (LT) to form coding sheets and categories freely generated and grouped through the abstraction process.14 The coding scheme was revised over a two-round iterative process of discussion and revision involving LT and DMM. Two coders (LT and SR) then independently applied the final coding scheme to the full list of questions. Discrepancies were resolved by discussion between LT and SR.

Demographic data

The five participating general practice groups had a total of 53 GP members. The number of years practicing per GP ranged from 1 to 43 years. All practices were classified as ‘urban’ on the basis of population ranges according to the Section of State Structure of the Australian Statistical Geography Standard.15 One practice was a university health service.

Description of the database

A total of 126 questions from the five participating practice groups were entered into the database over the period of 1.5 years (June 2016 to December 2017). The number of questions submitted per practice per month varied, with a range of 0–10 questions.

ICPC-2 PLUS

ICPC-2 chapter headings and clinical terms for coded questions are presented in table 1. Nine clinical questions were not able to be coded using ICPC-2 PLUS and hence were not included (eg, What are patient views and emotional reactions to health professionals looking up resources during consultations?). Of those that were coded, 20 questions fell under two of the chapter headings, and one fell under three chapter headings. Regarding clinical terms, 38 questions fell under two of clinical terms, and 11 fell under three or more.

Aside from the ‘general and unspecified’ chapter heading, the highest number of questions was coded to the ‘endocrine/metabolic and nutritional’ ICPC-2 chapter heading (n=19; 13.7%), followed by the ‘musculoskeletal’ chapter heading (n=16; 11.5%). Overall, the most common clinical terms were non-insulin-dependent diabetes (4%), blood test (endocrine/metabolic and nutritional (3%)); general and unspecified (3%)), and osteoporosis (3%). There did not appear to be large variations in the question types between GP practice groups although this was not checked for statistical significance.

Taxonomy of generic clinical questions

Using the taxonomy of generic clinical questions, the majority of questions were related to the broad category of treatment (71.3%), followed by diagnosis (14.5%), epidemiology (9.2%), management (2.3%), non-clinical (0.8%) and non-classified (0.8%). More specific generic question types are shown in table 2. 

Perceived knowledge gaps

Eleven categories of perceived personal knowledge gaps were identified from the inductive abstraction process (see table 3).

Summary of findings

This study of the types of questions submitted by 53 Australian GPs participating in an information service revealed that the majority of questions prioritised for answering by GPs related to treatment, followed by diagnosis and epidemiology, respectively. Using the ICPC-2 PLUS classification system, the most common chapter headings were ‘endocrine/metabolic and nutritional’ and ‘general and unspecified’, followed by ‘musculoskeletal’ and ‘digestive’. Through an inductive coding process, we identified several knowledge gaps; however, over 70% of all questions related to checking more specifically on one aspect of current evidence on a common problem, or to be informed about new tests or treatments and complementary and alternative therapies.

Several other studies have found questions concerning treatment to be the most common clinical question type.2 5 16 This may be unsurprising given the role of the GP to ‘treat all common medical conditions’,17 and patient expectations of (benefiting from) treatment.18 19 However, the percentage of questions pertaining to treatment was higher in our study compared with other Australian studies published 11 (35%)5 and 18 years ago (65%).6 This may be an artefact of slightly different classification systems across studies. For example, the taxonomy of generic clinical questions does not include a separate ‘prevention’ category, so may not reflect the expanding role of GPs in preventive healthcare activities. These include promoting smoking cessation, responsible alcohol consumption, weight control, physical activity and undertaking screening activities including blood pressure monitoring, cholesterol and blood glucose measurement and cancer screening. As well as generic question types, there are also some similarities between our findings regarding ICPC-2 PLUS classifications and other studies. Magrabi and colleagues, for example, found that gastrointestinal (13%), skin (12%) and musculoskeletal (11%) were the top three disease categories in a study of clinical questions from 227 Australian GPs,4 corresponding with two of three of our most common ICPC-2 PLUS chapter headings, and a study of questions submitted to a popular GP-restricted (Australia, New Zealand) Facebook group16 had two of the five major clinical topics (musculoskeletal and psychology) from our study. Future research may benefit from additional triangulation of data from different sources (eg, encounter observations; questions submitted to EBP question-answering services; questions submitted to public forums) to explore the role of different networks in answering clinical questions and to identify the most common question types across all forums.

Through the use of a standard classification system (ICPC-2 PLUS), our results can also be compared with findings from the Bettering the Evaluation of Care and Health (BEACH) programme; a continuous 18-year national study of general practice activity in Australia.7 BEACH data from 2006–2007 to 2015–2016 indicates that general and unspecified problems were the most frequently managed type of problem in 2015–2016; their management rate increased from 16.2 per 100 encounters in 2006–2007 to 20.0 per 100 in 2015–2016, an increase of 23.5% over the decade.7 There were also substantial increases in the management rates of endocrine and metabolic problems in general practice encounters between 2006–2007 and 2015–2016, from 12.1 to 13.5 per 100 encounters.7 In all years from 2006–2007 to 2015–2016, the five most commonly managed problems were hypertension, immunisation, upper respiratory tract infection, depression and diabetes.20 These mirror our findings and highlight that GP’s clinical questions overlap with the common problems they see in practice. Similarly, the categories for which we did not code any questions (eg, ear, eye and social problems) were in the six least common reasons for an encounter by ICPC-2 chapter, accounting for 3.2%, 1.9% and 1.0% of clinical questions from 2015 to 2016 BEACH data.7

A novel contribution of this study is the use of inductive coding of GPs’ questions to identify perceived knowledge gaps. The need to stay up-to-date with the state of evidence and be informed about new tests or treatments were the most prevalent knowledge gaps. Other studies have similarly identified the need to achieve basic ‘currency’ as a driver of clinical questions.21 Based on our inductive coding scheme, staying up-to-date included checking broadly on guidelines and recommendations for a common topic (10%) as well as checking in a more focused ‘specific’ way on one aspect of current evidence for a common problem (27%). Aligned with this, Brassil and colleagues found that more than half of the respondents stated that their clinical questions arose from ‘unusual cases’.22 This reinforces that even though GPs often manage common problems, focused questions about specific nuisances of a problem remain an important knowledge gap. In addition to achieving currency, 18% of questions in this study were coded as knowledge gaps about complementary and alternative therapies/medicines (CAM). CAM is estimated to be used by up to two out of three Australians,23 and other studies have likewise found a perceived lack of knowledge about complementary and alternative therapies among Australian GPs.24 In a study to assess questions about CAM to the Regional Medicines Information and Pharmacovigilance Centres in Norway, Schjott and Erdak found that 7.7% of all questions regarded CAM.25 Together, these findings highlight the growing prevalence of questions about CAM from mainstream clinical practice and the knowledge gaps which remain in this area.

Directions for research and practice

Our results highlight that GPs’ clinical questions frequently relate to treatment and to commonly managed problems, but are often focused more narrowly on specific aspects of a problem or new tests and treatments. The submission of such questions to our information service suggests that some GPs may not feel well-equipped to manage some of the diverse nuances of common presentations. In line with this, previous studies have found that published research and clinical guidelines do not align with the problems most frequently encountered by GPs21 and continuing professional development programmes do not always correlate well with the spectrum of common clinical issues experienced on a daily basis by clinicians.26

In Australia, the National Health and Medical Council recognises that, if guidelines are to be relevant, those who are expected to use them should play a part in their conception.27 However, there are currently no system-wide approaches for obtaining GPs’ evidence and knowledge gaps, and clinical practice guidelines are often commissioned by different groups in a top-down manner (eg, based on government health priority topics). Our system for ‘pulling’ real-time questions from GPs may help to circumvent this, and can inform the ‘push’ of more relevant and timely evidence for use in the clinical encounter. Other work has also shown that sustained relationships between clinical practices and academic groups can help to raise and answer clinical questions of importance to GPs.28 To ensure sustainability of this method to ‘pull’ clinical questions into evidence ecosystems, work is needed to support linkages between academic departments, general practices and guideline developers, as well as sustained and sufficient funding for infrastructure including staff.29

Our results also support the increasing advocacy for ‘living guidelines’.30 A living guideline is one that remains under review on an ongoing basis, with updates published at set intervals (eg, annually).25 The use of living guidelines can support clinicians to find evidence about new tests and treatments more easily and, if hosted on searchable digital platforms, may better support them to search for specific information about aspects of a problem.30 This represents an important advance on the lengthy, static clinical practice guidelines currently in use throughout Australia. Finally, our findings also reinforce the call for better awareness and resources about CAM. Even where there is limited or uncertain evidence for CAM, this should be included in clinical guidelines to ensure that clinicians are informed about the evidence and can communicate this to patients to support informed, shared decision-making. This may help to increase the number of people who consult their GP prior to using CAM (currently only 20% in Australia31) and is particularly important given the competing sources (social networks; books; media; internet31) and quality of information available to patients.

Strengths and limitations

We analysed questions from a small and non-randomly selected sample of GPs, which limits the generalisability of our findings. In fact, we purposively selected lead practices recognised as opinion leaders and early adopters of clinical advances. The fact that even these lead GPs had clinical questions which they requested assistance in answering supports the utility of, and need for, better systems for pulling real-time questions from GPs to guide research and the development and revision of practice guidelines. As all questions were submitted via an appointed GP liaison, we do not have information about which GP submitted each question, so some GPs may be over-represented in the database, and it is possible that the GPs actually contributing questions may be the most motivated and evidence-aware practitioners within those practices.

Direct observation studies such as that of Ely et al 3 have shown that GPs generate ~10–15 questions per day of practice. We, therefore, received a very select number of questions from participating GPs which is unlikely to be representative of the much larger number of questions that they generate. However, this study was designed to investigate the types of questions that GPs at the practice level have prioritised for answering to improve the quality of practice and uptake of evidence. The questions submitted to our service may also reflect those which GPs have particular difficulty answering through other means.

The use of multiple coding systems and independent coders are strengths of our study. Standard classification systems such as ICPC-2 plus allow for meaningful comparison of clinical questions over time and contexts. The ICPC-2 is now used in >45 countries as the standard for data classification in primary care,7 which also allows us to compare clinical questions to clinical presentations. We coded all questions submitted by GP groups, not only those which were answered by our information service.