Tests may miss what they are looking for or not alter management

For an investigation to be useful, it should have high sensitivity (the ability to correctly identify patients with a given disease) and specificity (the ability to correctly identify patients without said disease). Distinct but related clinically useful concepts are those of positive predictive value (the probability that someone with a positive test has the disease) and negative predictive value (the probability that someone with a negative test does not have the disease). It is of note that even when tests do have high reported sensitivity and specificity, there may be additional complexities in real-world settings. This is in part because when developing tests, figures of diagnostic accuracy are calculated in a specific context, which is often obviously symptomatic individuals compared with obviously healthy controls.

Many diagnostic investigations used by psychiatrists fail to meet these criteria, and many have such a low probability of a clinically relevant result that it is hard to justify their routine clinical use. Systematic reviews have found that routine blood tests rarely alter the management of patients due for admission to psychiatric hospital, even when abnormal results are found.^{Reference Chennapan, Mullinax, Anderson, Landau, Nordstrom and Seupaul6,Reference Conigliaro, Benabbas, Schnitzer, Janairo and Sinert7} In one of the included studies of 500 patients, only one individual had an abnormal result that mandated urgent medical intervention. Notably, this patient was obviously symptomatic.^{Reference Janiak and Atteberry8}

Imaging and physiological investigations may not fare better. EEGs show interictal epileptiform discharges (IEDs) in only 29–55% of people with epilepsy,^{Reference Pillai and Sperling9} so it is inappropriate as a rule-out test for complex partial seizures presenting with psychotic symptoms. In contrast, EEG abnormalities (broader than just IEDs) occur in 19% of those treated with antipsychotics, so an abnormal EEG is not specific to a neurological cause of psychiatric symptoms.^{Reference Centorrino, Price, Tuttle, Bahk, Hennen and Albert10} Similarly, one study found that 6% of healthy controls have at least one abnormality on an MRI of the brain, so the presence of an abnormality per se has a poor specificity for determining a neurological cause for psychiatric symptoms. This same research also concluded that none of 349 individuals with first-episode psychosis had MRI findings that necessitated a change in clinical management.^{Reference Falkenberg, Benetti, Raffin, Wuyts, Pettersson-Yeo and Dazzan11}

Investigations are no substitute for history taking and physical examination

Psychiatric patients invariably present in complex and myriad ways. In a minority, there is a recognisable physical cause for the psychiatric symptoms. For a physical illness to lead to psychiatric symptoms, there must either be a pathophysiological process (e.g. neoplastic, endocrine, autoimmune, metabolic, epileptiform, infective, neurodegenerative or metabolic) that is affecting the central nervous system and/or psychosocial mechanisms resulting from the physical illness. In each of these cases, it is very unlikely that a patient would present as entirely physically asymptomatic to an appropriately curious and observant clinician.

To put it another way, we might be surprised to see such physical illnesses presenting ‘occultly’, in the sense that a case would be missed after a thorough history and examination but picked up on a routine blood test or other untargeted investigation. Where psychosocial mechanisms predominate, these generally occur via experience of symptoms or via knowledge of a diagnosis, both of which can be elicited through an appropriate history or examination. In the specific case of functional disorders (such as functional neurological disorders), these are diagnosed through eliciting positive clinical signs, rather than being ‘diagnoses of exclusion’ which require investigation to exclude other causes.^{Reference Espay, Aybek, Carson, Edwards, Goldstein and Hallett12}

Even for tests with extraordinarily high sensitivity and specificity, there are further interpretative difficulties that arise in the absence of clinical suspicion. If a condition is very rare, then an abnormal investigation result may not be as diagnostic as is intuitively suspected. This is known as the ‘false-positive paradox’, or the ‘base rate paradox’, and occurs when the prevalence of a condition is lower than the test's false-positive rate (in these cases, the test will give more false positives than true positives). Crucially, the calculation is entirely different in populations who are already suspected to have the disease based on history and examination. In these cases, the pre-test probability (and thus the post-test probability after a positive test) would be much higher, and a resulting abnormal test much more clinically useful.

Take hypothetical blood test A which has a specificity of 99% for condition X. If condition X has a population prevalence of 1% and there are no additional features to increase the clinical suspicion of the condition (e.g. the patient has not been examined), we can assume the pre-test probability is 1%. In this case, an abnormal result from blood test A translates to a chance of having the condition (i.e. post-test probability) of only 50%. Most physicians, including us authors, would probably overestimate the post-test probability following a positive result on blood test A.

Investigations can be harmful

Investigations often have attendant risks, which are seldom adequately considered. In some circumstances, taking blood can be painful or uncomfortable (particularly in a distressed, paranoid or disoriented patient), risking injury for patient and clinician. The noise and enclosed environment of an MRI scanner can be frightening for an anxious patient. Over-investigation may also contribute to the development or perpetuation of health anxiety, as it can serve to confirm a patient's fears about a serious undiagnosed illness.^{Reference Howard and Wessely13} This may be particularly pertinent for individuals with somatoform or hypochondriacal disorders, who psychiatrists may encounter more often than might clinicians in other specialties.

A further important consideration is the impact of false positives. When tests are ordered in the absence of a clinical suspicion or hypothesis, a positive result inevitably returns the question: what next? In many cases, this might be further investigation, leading to a non-negligible potential of harm. Take the case of creatine kinase (CK), which is sometimes recommended as an admission blood test for psychiatric in-patients,² despite being incidentally and benignly raised in many patients, and even being a poor discriminator for neuroleptic malignant syndrome.^{Reference Chang, Harrison, Lee, Taylor and Stewart14} When faced with a raised CK in an asymptomatic patient, most of us feel that we should do something, so patients are frequently subjected to further blood tests and sometimes intravenous fluids for a test that should never have been done. More dramatically, if ten people are treated with endovascular coiling for an incidental unruptured intracranial aneurysm detected on MRI, one will end up disabled or dead within one year, all for a lesion that probably would not have caused any harm.^{Reference Al-Shahi Salman, Whiteley and Warlow15}

Readers may be familiar with ‘incidentalomas’, an all-encompassing term for non-specific lesions seen on medical imaging. Incidentalomas have dubious clinical relevance, but often prompt further investigations and cause anxiety to patients and professionals alike. Relatedly, the phenomenon of non-specific abnormal results also occurs with blood tests. One study of healthy adults receiving a routine battery of blood tests found that over one-third had at least one abnormal result, of which only 7% were deemed to require a medical review (none of them urgently).^{Reference Keuss, Parker, Lane, Hoskote, Shah and Cash16}

Finally, in the context of limited healthcare funding, expensive investigations divert resources from more worthwhile causes. Discounting the cost of staff time, a simple routine battery of blood tests is likely to cost in the region of £15–20.^{Reference Mughal, Narayanan, Gupta and Reay-Jones17} The number of psychiatric admissions needed to screen to find a positive serum free thyroxine (FT4) is in the region of 127, which corresponds to a cost of around £400 per abnormal result.^{Reference Arce-Cordon, Perez-Rodriguez, Baca-Baldomero, Oquendo and Baca-Garcia18} MRI brain scans cost in the region of £200. Cost-wise, there is likely more benefit from a five minute clinical history and examination.