Why do clinicians overtest? A narrative review and workshop group discussion exploring inuences on clinician overtesting as possible causes of overdiagnosis

Background: Screening and diagnostic tests provide important information to guide clinical management. Overtesting however may cause harm to patients and the healthcare system, including through overdiagnosis. Clinicians are ultimately responsible for test requests, and therefore are ideally positioned to prevent overtesting and overdiagnosis. Through this narrative literature review followed by group discussion at a conference workshop (Preventing Overdiagnosis Conference [PODC], Sydney, 2019), we aimed to identify and categorise factors that inuence clinicians to request potentially unnecessary screening and diagnostic tests that could result in overdiagnosis. Methods: Articles exploring factors affecting clinician test ordering behaviour were identied through a systematic search of MedLine in April 2019, forward and backward citation searches and content experts. Two authors screened abstract titles and abstracts, and two authors screened full text for inclusion. Identied factors were categorised into a preliminary framework which was subsequently presented at the PODC for iterative development. Results: The MedLine search yielded 542 articles; 55 were included. Another 10 articles identied by forward-backward citation and content experts were included, resulting in 65 articles in total. Following small group discussion with workshop participants, a revised thematic framework of factors was developed: their test ordering and assessing practices

Overtesting has become a growing concern in contemporary healthcare, [1][2][3] with worldwide movements such as the Choosing Wisely, Less Is More, and Too Much Medicine campaigns bringing the issue to the fore. [1,[4][5][6][7] Despite this, awareness and understanding of overtesting amongst the public, patients and clinicians remains limited. [3,8,9] Screening and diagnostic tests are integral to clinical management, whether they are routine blood pressure measurement, blood tests, imaging studies or more specialised investigations. However, overuse of screening and diagnostic tests can lead to harm through misdiagnosis, false positive results and false negative results. A less widely known harm from overtesting is overdiagnosis where people are labelled as having a "disease" for a condition that would not have caused them harm if it were left undetected and untreated. [1,5,6,[8][9][10][11][12] As well as harms from the disease label itself, overdiagnosis is usually associated with overtreatment, and may initiate a further cascade of unnecessary investigations and treatment. [1,10,11,[13][14][15] As such, using increasingly sensitive screening and diagnostic tests to detect more and earlier stage disease may sometimes cause more harm than good. [1,14] Individuals may be affected physically, psychologically and nancially [8][9][10][11]16] and the use of nite healthcare resources prevents their redeployment to others who would have bene ted from tests, treatment and other interventions. [8,10,11,16] A key way of preventing harms from medical tests in current use is to minimise overuse in people who are asymptomatic or at low risk of a clinical event or condition. [10] As "gatekeepers" [17] to accessing further healthcare, it is clinicians who are ultimately responsible for requesting screening and diagnostic tests, making diagnoses and offering treatment and/or further tests. Thus, clinicians are in an ideal position to limit overuse of tests and mitigate potential harms.
To date, there has been no systematic synthesis of published evidence to explore factors that in uence clinicians to request potentially unnecessary screening and diagnostic tests that may cause overdiagnosis. In this narrative literature review and workshop discussion with experts in preventing overdiagnosis, we aimed to identify and categorise important factors that may in uence clinicians to request unnecessary medical tests that may contribute to overdiagnosis. The results of this study will inform further planned research to develop a scale assessing clinician knowledge, attitudes and practices around overtesting and overdiagnosis, in order to measure the impact of clinician-targeted interventions to prevent overdiagnosis.

Methods
We conducted a narrative review of peer-reviewed literature using Boell's [18] hermeneutic approach characterised by a systematic but exible and iterative search strategy. A narrative review aims to summarise, interpret and critique available literature based on the authors' own understanding, or preexisting theories and models available. [19][20][21] Through an initial non-systematic search of the literature, we identi ed recurrent themes and developed inclusion and exclusion criteria and the search strategy.
We searched MedLine from inception to April 2019 for published articles using MeSH terms and keywords related to the concepts of "overtesting", "physician" and "attitude to health".[Supplementary File 1] We included articles that focussed on overtesting or overscreening and explored factors in uencing clinicians' test requesting behaviour. We did not limit our search to speci c conditions, tests or specialties. To supplement the MedLine search, we identi ed further articles through forward and backward citation and consultation with content experts in overdiagnosis. We excluded articles that focussed on overtreatment, unnecessary treatment or factors in uencing clinicians' treatment decisions.
Two authors (JL and KB) independently screened titles and abstracts. Where either author considered the paper potentially relevant, the full text was retrieved. Two authors screened the full text for inclusion in the review (JL screened all full text, and KB, FS and KP each screened one third). One author (JL) undertook inductive thematic analysis [22] of the included studies to identify distinct factors in uencing clinicians' decisions to order tests. These were initially categorised in a thematic framework of "direct/intrinsic" factors and "indirect/extrinsic" factors. A second author reviewed included studies and added additional factors to the framework (KB, FS and KP each reviewed one third of included studies).
[Supplementary File 2] The resulting framework of "direct/intrinsic" and "indirect/extrinsic" factors was presented at a workshop at the Preventing Overdiagnosis Conference in Sydney, December 2019. The workshop used small group discussion to receive feedback on the initial framework, and to elicit further factors that may in uence clinician test ordering behaviour not identi ed through the narrative review. Basic demographic information was collected from workshop participants with their consent. There were ve-six participants in each small group, with discussion facilitated by one of the authors. Discussions were audio-recorded and transcribed for analysis to further re ne the thematic framework developed in the narrative review.

Results of Narrative Review
A total of 542 articles were retrieved following the search of MedLine in April 2019. 55 articles were included after screening by title and abstract, and application of inclusion/exclusion criteria to full text articles. A further 10 articles identi ed through forward and backward citation and by content experts were included, resulting in a total of 65 articles in the review.[ Figure 1] The total number of articles reporting on each factor and quotes (from participants involved in qualitative interview studies) re ecting clinician attitudes and reactions are provided in [ Table 1] and [ Table 2]  [1,2,7,15,16,23-31,34,37-39,43−58] "You are so open for being sued by anything but it's very easy to want to lean towards the screening everyone… I de nitely think it's hard not to think legally" [38] "Once the issue has been raised, it is di cult to back away unless you are 100% because you are responsible if you are wrong" [38].
"I'm often a bit defensive...I guess that's partly that legal thing" [38] "I think the whole medical-legal thing also makes people more inclined to CT [computed tomography] someone even if they have a pretty low suspicion just 'cause no one wants to be sued" [39] "I think litigation is a problem; you miss one neck... fracture or bleed in the brain you are going to court" [47] Factor Articles  Illustrative quotes   Clinician  knowledge  and  understanding 25 articles [2,3,5,15,29,32,33,35,37-39,42−44,47,52,53,59−66] "How much work [laboratory testing] is, how much it costs, how much normal results can uctuate, things like that, I think we know very little about that" [3] "Nothing can really go wrong [with overutilization]" [3] "You understand the natural course of disease and the point in time at which you have to make a decision to do something different" [29] "When I'm admitting a patient or doing clinical work, it's kind of affected my thought process to where I think a little bit more about 'do I really need to get this test?', 'will it really change management?', 'could it potentially be harmful to the patient?'" [32] "Those like statistical issues don't apply to the individual...because...they make their decisions on a set of complex, but perhaps irrational basis, you know, anxiety and..." [38] "Yeah, so, I hate the D-dimer. I understand its utility. I think that too many D-dimers are sent... I think the decision to get a CTPA [computed tomography pulmonary angiogram] should be based on a clinician's clinical reasoning plus or minus the criteria, plus or minus a D-dimer" [39] "GPs may be playing a good game and saying I'm not going to bother this patient with having a GFR [glomerular ltration rate] of 59 because I know that although it quali es as CKD [chronic kidney disease] 3 it's not gonna make any difference to how I manage that patient and I think that's good medicine" [42] "As I said, a patient without previous medical history, without symptoms. In this case, I have never auscultated a lung and thought: "Thank god I listened to that lung." I mean, what do you expect from a healthy patient when you auscultate the lung? A healthy lung" [43] "When you have no idea what's going on, so it gives you something to hide behind" [47] "'Should be tailored according to family history, previous issues, lifestyle and previous ndings. Need to explain the limitation of check-ups" [65] Factor Articles Illustrative quotes Intolerance of uncertainty / risk aversion 22 articles [2,3,7,15,16,24,25,27,29,31,34,35,37,39,[43][44][45]49,50,53,67,68] "Lab testing is often only done for the doctor's peace of mind." [3] "I am worried if they don't have a full assessment and I miss something that it is going on with their heart that is not apparent because ECGs [electrocardiograms] and clinical examinations are not very precise" [29] "You're sitting there with someone who has a sudden-onset splitting headache, but otherwise you see nothing alarming… A CT scan for an acute headache. Even if the pre-test chance is 0.01. He does it anyhow. They have much more certainty than we do." [37] "What if it couldn't wait? How would you know it won't affect them?" [38] "You have to be self-con dent in not doing something" [43] Cognitive biases and experiences 12 articles [10,15,[28][29][30][36][37][38][39][40]46,69] "'There might be a bias to a situation where some doctors missed an important nding, when they were a junior doctor, so they always do scans because they are worried that something might happen like years ago" [29] "If you've ever experienced something like that, you can be sure that you'll send patients with vague complaints for further testing much faster. Absolutely" [37] "It's certainly a-hard to be, treating dying people who are young and not to worry about all of this and I, but I try not to change my practice based on my own personal experience of one or two people dying of prostate cancer" [38] "I would say that my clinical experience highly in-uences my ordering… sometimes I feel a certain way about a patient even though they don't t a certain pro le and I'll end up doing something additional for them" [39] "The initial thing was PSA [prostate speci c antigen] is useful and that has basically stuck in my head, that PSA testing is useful" [46] Factor Articles Illustrative quotes Sense of medical obligation 9 articles [1,7,11,37-39,41−43] "Some GPs mentioned their frustration at not being able to offer the patient something useful, at the feeling of empty hands, owing to the lack of a diagnostic or therapeutic plan for patients presenting with unexplained complaints. A test request symbolises a serious attempt to deal with the patient's complaint" [37] "To not screen somebody, I don't know, it seems cruel, it's cruel and irresponsible... to not at least make an attempt to avoid the misery of a person getting prostate cancer, to me, seems unbelievably cruel" [38] "We have to diagnose them if they have a problem" [38] "If it's on your radar… you're almost honorbound to do the study of choice" [39] "My personal policy I would always disclose...generally speaking I would always explain the diagnosis" [42] "'Action' dogma of doing anything possible for the individual patient" [43]  "It can reduce the anxiety and prevent representations to the hospital, helping to keep them from coming in with chest pains" [29] "Now she had problems with her feet and arms, morning stiffness, pain in the joints. But there was no redness, no swelling, wasn't warm, functioning was good. But she was still uneasy. I had to con rm this to her with a blood test, otherwise the discussion would go on and on" [37] "But the GP lives in the community, has to continue caring for the patient. If you really mess things up, so that the patient switches to another doctor, that's what affects me" [37] ""So they see it as their right to have it" [46] "I guess I do it because...I want my patients to perceive that I practice good medicine...you do have to be seen to be proactive" [46] "There is a demand from patients for testing or medication or imaging that they've read about or they feel that they should get in order to be satis ed that they've been adequately cared for" [47] "Patients absolutely drive test ordering..." [47] "Patients come in and they say, 'Oh, I have this, and I want a CT scan done.' They'll tell you what they want done" [51] "If we order more tests and we make sure we have every test ordered that might possibly be needed, the patient's happy and leaves in their ED [emergency department] stay" [51] "I'll say "well you just had one two years ago, you're on treatment, it was stable from the year before, and I don't think you need one"… what does usually happen is that they usually win" [64] "Can improve relationship between patients and doctor" [65] "Check-ups are largely patient driven secondary to media/public health generated anxiety" [65] Factor Articles Illustrative quotes Guidelines, protocols and policies 21 articles [25,28,29,43,44,39,38,46,42,47,50-52,64,66,68,78,72−75] "There are situations where I've ordered an echo when I otherwise would not have because guidelines mandated" [29] "I think there's more, as much as we've developed these decision rules-I think there's a lot to be said about just experience" [39] "I think people are wary of practicing not in line with that and then they have potential then for criticism" [46] "There's plenty of guidelines, but they're all different and there's nothing o cial...there's no hard and fast rule" [46] "Because I work in a teaching practice, my residents are very devoted to guidelines. A lot of them are driven by the more recent guidelines" [64] Financial incentives "Identifying more disease means more business" [15] "If I went around having my 10 minute discussion with all my patients about why not to do PSA testing, I will make less money than [a GP] who does the 30 second -here Jack, that's a good idea, here, have the PSA test" [46] "To be perfectly honest, I only do it because of patient expectation as a business decision, not as valid evidence based medicine." [65] "A lucrative source for the private hospitals" [65] Factor Articles Illustrative quotes Pressure from colleagues (and medical culture) ‡ 13 articles [3,28,29,37,38,43,45,46,51,68,73,76,78] "Well, often the supervisor just says to run some tests, and I just accept that without question" [3] "I recently ordered a lipase, but then the gastroenterologist called me and said: in this hospital, we always combine it with an amylase." [3] "If an experienced cardiology colleague says we should do another echo, I would not feel strong enough to say no" [29] "If the neurologist had written, "There's nothing the matter" ... But how must I say "you have to accept it" if the neurologist says that perhaps the patient should be looked at by someone else" [37] "If I get a letter from the diagnostic centre with the comment "You request 10% more than the average GP in Maastricht", then you get critical. You wonder if we should wait a bit longer with this patient" [37] "If you're not going to order it, the next doctor will" [51] "He would see the cardiologist every three months and would get a stress test every year...When he came to see me...I had to tell him 'I don't think that that's necessary" [51] "A lot of tests get done that probably don't need to get done because our residents are afraid of not ordering something because they'll disappoint us" [51] Factor Articles Illustrative quotes Time constraints, (physical vulnerabilities and language barriers) † 13 articles [3,16,26,29,37,39,44,47,49,51,57,65,68] "Some days patients want tests that I feel are not necessary but I want to avoid discussions or I'm tired and I will order tests anyway" [29] "You see many exams ordered, "Rule out PE [pulmonary embolus]," and that's all that you have… we often just go ahead and do the exam, to be honest, because it ends up creating a lot of lost time" [39] "If you had enough time to do a thorough historytaking of all these people… People would say '"I think I've been well understood, listened to, and examined", and need far fewer further investigations. But that is much too time consuming" [37] "If I'm really busy and I have ten people in the waiting room, and if I feel pressured and overwhelmed, I can say,'Yep, here is a requisition for the MRI [magnetic resonance imaging], let's get it done and move along." [47] ""They do a lot of catscans because they don't have time to observe patients… work them up, get them out the door." [51] "A major concern that it could increase workload which would diminish time for treating ill patients" [65] Availability, ownership and ease of access to tests 11 articles [3,15,28,29,30,44,39,37,69,75,77] "Checking boxes on the lab form, I often go, let's do this one too, and that one" [3] "When you're ordering lab tests, it is easy to just order some more tests" [3] "The patient is already being sent for another test to the diagnostic centre, which creates a low threshold for doing more testing… so why not?" [37] "I think for any test if it's very, very available and it's fast and it's easy to do and it doesn't take a lot of time and there's more turnaround on the report-then we're just more likely to use it more." [39] "It would probably be valuable to make the process less convenient because the threshold is so low to order CTs" [39] Factor Articles Illustrative quotes Pre-emptive testing for subsequent care 10 articles [26,29,43,46,47,51,[63][64][65]73] "I am glad that I can refer to something… And you could describe that as medical overuse to some extent. Because we are talking about tests which were not totally urgent or rather luxurious given the speci c symptoms at that time. But it can be really helpful to have this reference point" [43] "People are used to sort of being screened...so we're tacking this onto the discussion basically" [46] "They will tend to steer on the side of getting a test, even though it may be unnecessary, because they fear they will not be able to get the patient referred" [47] ""We order tests because we feel we have to get everything up front, because it's just too painful to do things too slow, to do things as a series" [51] "Often I'm doing [BMD tests] at menopause time in a woman's life when things sort of come up. I get a baseline maybe at menopause" [64] "You only realize the importance once you do it-the yield of signi cant results is surprising" [65] Contemporary medical practice and new technology 5 articles [7,28,29,43,77] "There is less emphasis on clinical examination. Nowadays we hear murmurs, and we try to quantify their severity which leads straight to ordering an echo… However, this can result in overuse of imaging" [29] "The greatest challenge will be to put more emphasis on history taking and physical examination again… This is the prerequisite to avoid further unnecessary investigations" [43] Following expert focus group discussion: † "medical culture" was grouped with "pressure from colleagues" ‡ "physical vulnerabilities" and "language barriers" were grouped with "time constraints" Direct/Intrinsic Factors

Fear of malpractice and litigation
Thirty-four articles cited fear of malpractice and litigation as a cause for unnecessary testing and diagnosis. A number of studies demonstrated a positive correlation between likelihood of ordering imaging tests and level of litigation concern.[23-26] Hoffman and Kanzaria [27] referred to a survey of United States (US) emergency physicians where 97% of respondents ordered advanced imaging tests due to fear of litigation and missing a low probability diagnosis despite feeling the tests were unnecessary. Sanabria et al [28] showed that pathologists concerned about litigation tended to lower implicit disease thresholds for indeterminate and malignant tumour diagnoses. Conversely, a number of studies deemed malpractice not to be a major driving factor for ordering tests. [29][30][31] Clinician knowledge and understanding Twenty-ve articles highlighted unnecessary testing as a compensatory measure for lack of knowledge and understanding. For most studies, this related to lack of knowledge and understanding of the drivers of overuse, the natural history of disease, and appropriate management pathways. Cardiologists and non-cardiologists [29] reported that greater training and experience helped them understand when to perform an echocardiogram and in which patients as they "understand the natural course of disease". [29] Through their novel re ective writing program, Caverly et al [32] demonstrated that greater cognisance of the drivers of overuse can positively impact clinical decision making and test ordering behaviour through empowerment to identify, discuss and avoid overuse.
Wegwarth et al's [33] randomised trial assessing 412 primary care physicians' understanding of common screening statistics found that lack of knowledge and understanding of signi cance of test properties and results also encourages overtesting. Physicians were more likely to recommend a test when presented with evidence based on increased 5 yearly survival rates (which tend to overestimate the bene t of early detection and treatment because of lead time bias and overdiagnosis) as compared with reduced mortality rates (which give an unbiased estimate of bene ts). [33] Intolerance of uncertainty and risk aversion This factor was cited in 22 articles. In the 2017 American Board of Internal Medicine Foundation survey, [34] among the commonest reasons for ordering low value tests were desire to reduce uncertainty (84%) and "just to be safe" (78%). [34] Egerton-Warburton et al [35] found that over half of 1029 emergency physicians in the study reduced their implicit "test threshold" well below the explicit threshold set using empirical data on test accuracy and risk of harm from the disease, due to their discomfort with diagnostic uncertainty. [35] Coon et al [16] also highlighted the "shotgun approach" [16] of ordering a broad range of tests and hoping for a positive result somewhere, which is often used in situations of diagnostic uncertainty. However, in assessing factors in uencing tendency to order imaging using hypothetical scenarios, Kini et al's [25] survey of cardiologists and general practitioners did not nd a signi cant association between risk aversion and tendency to order cardiac stress tests and echocardiography (likely due to their small sample size). [25] Cognitive biases and previous experiences Unnecessary testing often occurs as a result of cognitive biases; three signi cant cognitive biases were found in 12 articles. The rst was "availability bias" [4,38] which occurs when the likelihood of future events is estimated based on ease of recall of similar events. Closely associated is the impact of previous experience of clinical events. Clinicians with recent negative experiences or "recent medical blunders" [37] were more likely to adopt an aggressive approach to diagnostic testing. [10,30,37,38] Gyftopoulos et al [39] and Sanabria et al [28] also suggested that positive experiences from test ordering can increase the likelihood of ordering further tests in the future.
The second was the "fuzzy trace theory" [36] or "gist memory" [36] which occurs when perceptions are based on a general impression of information rather than speci c details. In the setting of test ordering, tests are more likely to be ordered when there is an greater perceived bene t and less likely when there is greater perceived harm. This was seen in Elstad et al's [36] cross-sectional study of 126 clinicians who, after being presented with two hypothetical clinical vignettes, were more likely to recommend colonoscopy due to greater perceived bene t than harm and were less likely to recommend prostate speci c antigen testing due to greater perceived harm than bene t.
The third cognitive error is "representative bias" [10,15,40] or "base rate neglect" [10,15,40] which occurs when failing to take into account pre-test probability in estimating post-test probability in the setting of a positive test result. [10,15,40] By overinterpreting positive results, clinicians are more likely to order further tests. In Austin's [40] survey, it was found that 10% of physicians had incorrectly deemed positive predictive value to be the same between screening (where pre-test probability is low) and diagnostic (where pre-test probability is high) tests.

Sense of medical obligation
This factor was cited in 9 articles. Testing based on a sense of medical obligation was driven by the psychology of regret [1] in missing serious disease or making a diagnosis too late [7], and way of showing that everything was being done for the patient. [23] A US survey [41] of primary care providers' and gynaecologists' breast cancer screening practices showed that those with greater levels of anticipated regret were more likely to recommend mammography. [41] Simmonds et al showed that when faced with the decision as to whether to disclose a diagnosis of clinically inconsequential chronic kidney disease, some general practitioners felt morally bound to respect patient autonomy and disclose so as to allow them to make their own health decisions. [42] Indirect/Extrinsic Factors

Pressure from patients and doctor-patient relationship
This was reported in 29 articles. Van der Weijden et al [37] noted that not only were anxious patients more likely to request tests, but general practitioners conscious of long term relationships with patients were more likely to meet these requests. In the 2014 and 2017 American Board of Internal Medicine survey [34], a large proportion of participants cited desire to keep patients happy, patient's insistence, and the idea that patients should make the nal decision, as reasons for ordering tests. Gogineni et al [70] and Gri th et al [71] showed that clinicians were more likely to acquiesce to patient demands for tests the clinician judged as unnecessary if patients threatened to see another clinician. Conversely, He [30] and Siedlikowski et al [68] found that the better the relationship, the less likely patients were to demand unnecessary tests, and the less likely doctors would be to order unnecessary tests.

Guidelines, protocols and policies
Interpretations and attitudes toward guidelines, protocols and policies were found to signi cantly in uence test ordering behaviour in 21 articles. Akerman et al [72] found that that there was a drop in prostate screening rates from 91.7-80.4% when new recommendations from the Canadian and United States Preventive Services Task Force were released following evidence of little net bene t from screening. A number of studies showed that non-existent or discordant guidelines can result in overtesting. [38,68,[73][74][75] However, some studies demonstrated that the use of protocols resulted in inappropriate use of tests and overtesting. [28,29,66] Alber et al [43] and Bishop et al [51] found that doctors based their test ordering decisions on individual patient cases, using guidelines "as a guide rather than strict rules". [51] This resulted in variation in test ordering behaviour, with some more likely to test while others less likely. [43,51] Financial incentives This was reported in 20 articles. Moynihan and Doust [1] and Sanabria et al [28] refer to the concept of "physician induced demand" [15] whereby physicians can order return visits and perform diagnostic tests when indications are vague or controversial. Pickles et al [46] showed that Australian general practitioners (fee-for-service health system) were more likely to order prostate speci c antigen tests than their United Kingdom counterparts (no fee-for-service health system). Fonesca et al [29] suggested that clinicians working in private healthcare settings were more likely to be driven by economic incentives in regards to test ordering. Physicians who own imaging equipment engage in more testing with similar clinical outcomes, [10] implying a nancial con ict of interest as a driver of excessive testing. Conversely, general practitioners in Simmonds et al's [42] study were resistant to being incentivised to keep a register and monitor patients with stage 3 chronic kidney disease due to con ict with their beliefs around the meaning of mild reduction in kidney function and belief that "they're creating an illness that doesn't exist". [42] Pressure from colleagues Pressure from other clinicians was as a driver of unnecessary testing in 13 articles. Fonesca et al's [29] cardiologists admitted to ordering echocardiograms when colleagues deemed it necessary despite themselves thinking otherwise. D'Souza et al [76] also showed that junior doctor test ordering was signi cantly in uenced by their peers, colleagues and supervisors. Siedokowski et al [68] found that as many as 89.6% of physicians would order a screening test they would not have otherwise ordered if specialists had recommended the test.

Time constraints
Time constraints for clinical assessment was another factor behind excessive testing in 13 articles. A number of studies recognised the time pressures in the work environment which limits time with patients and encourages physicians to provide a test just to expedite the clinical encounter. [26,37,47,51,57] Ellen and Horowitz's [44] survey of Israeli nurses showed that more than half felt that giving physicians more time to discuss alternative tests would reduce overuse. On the other hand, Murphy et al [65] highlighted that some general practitioners felt doing an annual general check-up on otherwise healthy patients represented a waste of their limited time that was better spent attending to sick patients.
Availability, ownership and ease of access to tests This factor was reported in 11 articles. Tests were more likely to be ordered when logistically easier, for example, during day shifts as opposed to evening or night shift, [44] when in closer proximity, [15] when there was little resistance in test ordering [39] or when able to order from desktop devices. [37] Fonesca et al [29] also noted that waiting times and patient physical mobility affected likelihood of echocardiogram test ordering. Ownership or having the available technology and equipment was also a key predictor of testing ordering. This was the case in tertiary hospitals [30] and amongst those who owned a CT scanner and used it for prostate cancer staging and breast cancer surveillance. [69] Pre-emptive testing for subsequent care Reported in 10 articles, uncertainty about what future tests would be required, a desire to avoid delays in a patient's care pathway and testing to establish a "baseline" were recognised as causes for potentially unnecessary testing. Amongst Irish interns, Flynn et al [73] noted that almost half requested tests they felt their consultant would want, to ensure that cases were never cancelled due to a lack of data. Sears et al [26] found that 76% of physicians felt they couldn't refer a patient to a specialist without having magnetic resonance imaging done rst. [26] Similarly, Alber et al [43] highlighted that some general practitioners considered "medical overuse in inpatient care as a welcome diagnostic work-up and baseline for the subsequent outpatient care" [43] with Munce et al noting that some family physicians ordered bone mineral density tests in asymptomatic women at menopause to obtain a "baseline". [64] Contemporary medical practice and new technology Four articles highlighted that in contemporary medical practice, history taking and physical examination are low priority with emphasis being placed and even reliance on technical tests, which contributes to overtesting. [7,28,29,43] Lysdahl and Hoffman [74] also showed that improved radiological technology was a major cause of overall increased investigation volume that, in turn, predisposed to unnecessary investigation.

Results of Expert Group Discussion
There were 15 participants at the Preventing Overdiagnosis workshop. All but one participant were either academics/researchers or medical doctors and most had been working for at least ve years.
[Supplementary File 3] Participants generally agreed with the importance of factors identi ed through the narrative review. A number of additional suggestions were made and subsequently integrated into the existing framework.
Physical vulnerabilities (such as being fatigued or hungry) and language barriers were included in "time constraints" as participants felt that such motivational factors for test ordering were based on the desire to reduce patient-contact time. The notion of wanting to t into the existing medical culture was included in "pressure from colleagues".
Participants believed that the initial framework of "direct/intrinsic" and "indirect/extrinsic" factors was too dichotomous and did not adequately illustrate the overlap and interaction between all the factors.
Participants considered that most of what was conceptualised as "indirect/extrinsic" is actually moderated by the "direct/intrinsic" values and perceptions of the individual clinician, and that neither of these properties are xed but change over time. Participants suggested modifying the proposed framework to better highlight the complex relationship between factors in uencing clinician test ordering behaviour.
Based on this feedback, we revised the initial framework to incorporate the factors in a new framework of: "interpersonal", "workplace environment", "broader context" and "intrapersonal" factors.[ Figure 2] "Interpersonal" factors are those related to direct interactions between a clinician and other clinicians or patients. "Workplace environment" factors refers to those potentially encountered on a day-to-day basis in clinical practice whereas "broader context" factors refers to factors that indirectly or subconsciously

Discussion
This narrative literature review, and subsequent small group discussion with experts in preventing overdiagnosis, highlights the myriad of factors that in uence a clinician's decision to order screening and diagnostic tests. We present a framework for understanding factors that in uence clinicians to request potentially unnecessary tests that may cause overdiagnosis. The most commonly cited factors were "fear of malpractice and litigation", "pressure from patients and doctor-patient relationship", "clinician knowledge and understanding", "guidelines, protocols and policies" and "intolerance of uncertainty and risk aversion". These were grouped within a nal framework of "interpersonal", "workplace environment", "broader context" and "intrapersonal" factors.
Physicians' fear of malpractice has been cited as the primary cause of medical excess [45,80]. With societies becoming increasingly punitive [48] for errors of omission rather than commission, [16] clinicians are understandably motivated to order tests as a strategy for lowering legal risk and avoiding the nancial and emotional consequences of litigation. [27] However, this perceived risk is likely exaggerated and largely unfounded. [58] To be subject to a lawsuit, clinicians must act well below the standard of care of their peers, and this usually occurs where there are poor doctor-patient relationships or communication [58] Missing a diagnosis or having an undesired outcome after deciding to not order a test based on sound clinical reasoning therefore falls well short of litigation. Legal reforms to reduce clinician liability have done little to deter unnecessary testing. [27,81] For example, in New Zealand where there is a no fault system, inappropriate use of medical resources still takes place. [82] Increasing clinician understanding of actual risk for lawsuits, and ways to decrease their personal risk such as using open communication [83]and shared decision making [58] may reduce overtesting driven by legal liability fears.
Pressures and demands from patients is a growing issue in modern healthcare [70] partly driven by information publicly available on the Internet [29,71] and social media. [15,46] Such information may be low quality, unreliable or incomplete, [71] and patients may have unrealistic or ill-informed expectations.
[84] Patient demand, whether clinically relevant or not, is typically driven by the "worried well" [85] or by those with a distrustful relationship with their clinician. [30,68,70] Moreover, doctors working in private practice are more likely to succumb to these pressures out of a desire to maintain a good doctor-patient relationship and ongoing business. [29,44] Clinicians may be better equipped to respond to inappropriate patient demands (while still maintaining a good relationship) with targeted communication skills training emphasising clinician-patient interaction. [29,79] Clinician understanding and awareness of natural history of disease, disease management, and test bene ts and harms is integral to preventing overdiagnosis. However many clinicians have been found to overestimate bene ts and underestimate harms, [60] lack ability to retrieve relevant evidence, [60] or have poor statistical numeracy and risk literacy. [66] This makes it di cult for clinicians to not only understand medical test statistics, but also to communicate this information to patients. A greater emphasis on interpretation of statistical information in medical school curricula could equip clinicians with these essential skills. In the clinic, decision aids support clinicians to understand and explain complex information. [86] Guidelines, policies and protocols can prompt clinicians to both limit testing and overtest. An example of the positive impact of guidelines in reducing overuse has been with prostate cancer screening. [72] Conversely, clinicians may order tests that are not clinically indicated because policies mandate it. [29] Guidelines are considered a reputable source of evidence based recommendations to support clinical decision making. [87] However, they can also be marred by vested interests, integration of poor quality or outdated studies and have limited applicability to individual patients. [43] As such, guidelines, protocols and policies should only be followed when after rigorous evaluation of quality and validity.
Intolerance of uncertainty and risk aversion is regarded as the most common reason physicians engage in medical excess. [27] The culture of "shame and blame" [27] and medical education that instils fear of uncertainty [16] only exacerbates the situation. A number of studies attribute poor medical training and lack of experience to higher levels of uncertainty. [43,53] This is in contrast to Fonesca et al [29] who suggests that even experienced specialists may feel the need to reduce clinical uncertainty through test ordering. Regardless, different individuals will have different innate aversions to risk and error.
Uncertainty will always exist in medicine [48] and continually performing more tests to allay these uncertainties will only result in the diversion of scarce resources to where it is not needed. By emphasising their prioritisation of patient's best interests, engaging in shared decision making and communicating this inherent uncertainty with patients, clinicians can feel empowered to deal with such uncertainty. [88] Strengths of our study are the systematic approach to retrieving the literature, use of at least two authors at each step of the review process and rigour in thematic analysis of the data with multiple iterations through discussion with all authors. We were able to engage with content experts at the PODC workshop, and revise the initial proposed framework. Our nal framework is not only more useful for measuring clinician-targeted strategies and interventions, but also explores the complex interplay of factors that occurs in real clinical practice. A limitation of our study was that we searched only one database which may have limited the number of articles included. However, given the richness of data and complexity of the thematic framework, it seems unlikely that we missed important factors that in uence test ordering behaviour. This is also supported by the fact that the small group discussion did not generate any new, distinct factor groups.
We are aware of no other systematic evidence synthesis speci cally exploring the factors that in uence clinician decisions to order unnecessary tests. Pathirana's [7] analysis article explored the drivers of overdiagnosis and mapped them to potential solutions but did not focus on the drivers from a clinician perspective. Siedlikowski et al [68]and Sharma et al [75] identi ed a number of factors in uencing clinicians' recommendations for mammography screening, but not test requesting behaviour more broadly.

Conclusions
Although medical tests are integral to clinical management, clinicians have the ability and responsibility to limit overuse of tests to prevent harms to patients and the healthcare system, particularly those arising from overdiagnosis. By using this thematic framework of interpersonal, broader context, workplace environment and intrapersonal factors in uencing decisions to order tests, clinicians can become more aware of their test ordering behaviour and take steps towards protecting the welfare of individuals and promoting more sustainable healthcare.
The results of this study will be used to inform the development of a scale to assess clinician knowledge, attitudes and practices around overdiagnosis. In this way we hope to measure the effects of cliniciantargeted interventions to prevent overdiagnosis. Form for their contribution to be audio-recorded and their responses to the Demographic Questionnaire to be used as data for analysis in the research project (forms available upon request).
Availability of data and materials: The datasets generated and/or analysed during the current study are available in the Dryad repository. The authors declare that they have no competing interests Funding: Figure 1 Study ow diagram