The Role of Bias in Clinical Decision-Making of People with Serious Mental Illness and Medical Co-morbidities: a Scoping Review

Abstract

The aim of this review was to examine the evidence for the impact of explicit and implicit biases against mental illness on the clinical decision-making of primary care physicians, medical students, and nurses when they are providing care to individuals with serious mental illness for cardiovascular disease, diabetes, and cancer. Studies were identified by searching MEDLINE, EBSCO host, and PsychINFO. A total of 18 studies published between 1996 and 2020 were reviewed and summarized. The studies were divided into two groups—studies that used a simulation or vignette methodology and those with a qualitative approach (interviews and focus groups). Of the simulation/vignette studies that allowed participants to report what they would have done in various clinical scenarios, there were roughly equal numbers of neutral or negative clinical decisions that represented 80% of the relevant behavioral results. Only 21% of the findings demonstrated a clinical decision that was favorable towards people with mental illness. Of the qualitative studies, all of the studies reported behaviors (either self-reported or observed) that were likely to be biased against people with mental illness, while 3 of the studies reported mixed results. Healthcare provider bias against individuals with mental illness does exist and impacts clinical decisions negatively. Much more empirical work needs to be done to determine the full extent and impact of the problem, including how these decisions affect the lives of individuals with mental illness.

Introduction

Previous work has found that individuals with serious mental illnesses (SMI), such as schizophrenia, bipolar disorder, and major depressive disorder, have more medical comorbidities and die younger than peers without SMI.^1–4 Factors such as risky personal habits (e.g., smoking and limited exercise), side effects of psychiatric medications, socioeconomic issues,⁵ and shared pathophysiology that increases the risk of development of physical disease,⁶ all impact the higher morbidity and mortality of persons with SMI. Disparities in access and delivery of healthcare may also play a role.⁷

Medical care that people with SMI receive for major physical illnesses such as cardiovascular disease (CVD), diabetes, and cancer has also been shown to differ from that received by people without SMI.^8–15 A recent meta-analysis identified disparities in screening and interventions for CVD, including catheterization, revascularization, and medication (except when a mood disorder was present).¹² Similarly, people with mental illness are less likely to receive all recommended services and care support for diabetes,⁹ including being hospitalized less frequently.¹⁴ A review of cancer treatment found that fewer patients with SMI received proper treatment, resulting in a lower survival rate, and that end of life care was also lacking.¹⁵

Many explanations for these disparities have been proposed in the literature, including the idea that clinician bias against people with SMI may affect clinical decision-making.^8,9,11 Prior reviews have found evidence of such clinician biases;^16,17 however, most studies do not have data to support a causal link between biases and clinical decision-making. The existence of negative beliefs or attitudes towards persons with SMI in clinicians does not necessarily equate to a significant impact on clinical behavior, given that explicit attitudes and behaviors are often only weakly related.^18,19 Self-reports of attitudes and performance of behaviors can be heavily influenced by context, most notably social desirability demands.²⁰ For example, clinicians participating in studies may consciously choose not to report negative biases and may override discriminatory behavioral impulses that align with negative biases in order to be perceived more favorably.²¹

Research has examined explicit biases in mental health professionals. Mental health professionals have a complex relationship with people with mental health disorders. In their role as mental health providers, they often suffer from stigma and prejudice themselves by virtue of who they care for.¹⁷ But they can also be stigmatizers, and studies have been mixed as to their explicit beliefs about mental illness, people with mental illness, and treatment of mental illness.¹⁷ Fewer studies have examined explicit biases against individuals with mental illness in non-mental health care professionals. In one review, stigmatizing attitudes towards people with mental illness and a desire for social distance and unease were found in a variety of non-mental healthcare professionals.¹⁶ These attitudes in turn were linked to avoidance and rejection of people who are seen as difficult to treat. Fear and discomfort were also reported and thought to be due to a lack of knowledge about mental illness by medical providers.²² Primary care providers have been noted to feel unprepared to deliver appropriate care to patients with mental illness.²³ Physicians and nurses have also reported explicitly negative and biased views of individuals with SMI that include beliefs that such individuals are violent, manipulative, have a lower likelihood of treatment adherence, and are poorly motivated to change.^17,24

Implicit attitudes (i.e., largely unconscious, unreportable attitudes that are typically measured via indirect means, such as the Implicit Association Test, or IAT)²⁵ may also predict behaviors, particularly spontaneous o nes¹⁸ and can be influenced by situational context, including mood and self-regulatory resources as well as task instructions.²⁶ There have been studies demonstrating negative implicit bias against mental illness in resident p hysicians^27,28 and other healthcare providers.²⁹ One study showed that implicit attitudes predicted over-diagnosis (i.e., assigning diagnoses beyond the correct one) of mental illnesses from providers with varying levels of mental illness training.³⁰ However, as with explicit attitudes, studies causally linking implicit attitudes to clinical behavior are lacking.

Given the increased mortality in persons with SMI,^1–4 it is imperative that any link between provider bias and clinical decisions be clarified. The purpose of this scoping review is to examine the evidence for the impact of explicit and implicit biases against mental illness on the clinical decision-making of primary care physicians, medical students/residents, and nurses when they are providing care to individuals with serious mental illness (major depressive disorder, bipolar disorder, schizophrenia, schizoaffective, or other psychotic disorders) for cardiovascular disease, diabetes, and cancer.

Methods

PubMed/Medline, EBSCO host, and PsychINFO were searched on August 16, 2019. The search strategies were formulated by author RH in consultation with a medical librarian. The PubMed search string (Appendix 1) consisted of MeSH terms and keywords for each component of the research question, with the exception of the outcome. Terms were chosen to identify serious mental illnesses (major depressive disorder, bipolar disorder, schizophrenia, and schizoaffective disorder primarily), medical illnesses (cardiovascular disease, diabetes, and cancer), and provider attitudes/biases. Outcome was excluded due to the variety of terms that could be considered “clinical decisions.” Instead, clinical decision-making was incorporated in the inclusion/exclusion criteria. The search string was adapted for PsychINFO (see Appendix 2). The initial searches returned 1784 and 2118 articles, respectively. However, two of the 21 starting references were not picked up by these searches. Two steps were taken to expand the search. First, each of the 21 references was searched for in PubMed, and then, the “similar articles” list provided was downloaded. Each article had somewhere between 70 and 150 “similar articles” listed. Secondly, the “cited by” list provided by PubMed for each of the 21 references was also downloaded. These methods yielded 4038 articles to be screened.

All search results (N = 7961) were uploaded into Covidence^™ systematic review management software. After removing 1083 duplicate records, 6878 titles and abstracts were screened, 298 of which moved on to full-text screening (PRISMA diagram in Fig. 1). Inclusion criteria were as follows: (1) full text accessible, (2) peer-reviewed journal articles, (3) published in English (4) after January 1990, (5) containing primary research using qualitative, quantitative, or mixed methods (6) to examine the impact of attitudes about mental health (explicit or implicit) held by primary care physicians, medical students, or nurses on clinical decisions made in regard to treatment for physical illness in patients with serious mental illness. In the end, 16 articles from the search (5 qualitative, 11 quantitative) met all inclusion criteria. The reference lists of these articles were screened for other articles potentially appropriate for inclusion. A total of twenty-four references with promising titles were examined but did not meet inclusion criteria.

Fig. 1.

PRISMA diagram

The search was repeated on 03/01/2022, with only search results published between the original search date and the repeat search date examined for inclusion. Through the same process, two additional articles,^31,32 out of 1030 examined, were determined to be eligible for inclusion for a total of 18 included articles.

After article selection, data extraction was performed in Covidence. Each article was read by two team members who independently abstracted the study details. Next, two team members independently assessed the quality of each article. Qualitative articles were assessed using the nine-item tool that evaluates the methodological rigor of a study with qualitative or mixed methods from Hawker et al.³³ Each of the 9 items (evaluating the quality of the abstract, the introduction, the methods and data, sampling, data analysis, ethics and bias, results, transferability/generalizability, and implications/usefulness) was rated on a four-point scale from very poor (1) to good (4); thus, the maximum quality score was 36 and the minimum score was 9. Scores were then categorized as high (or “good” per the instrument) (30–36), medium (24–29), or low (“poor” and “very poor” per the instrument) (9–24) quality. Quantitative articles were assessed using the Quality Assessment Tool for Quantitative Studies.³⁴ This tool evaluates articles across 6 domains (evaluation of potential selection bias, study design, confounders, blinding, data collection methods, withdrawals and drop outs, intervention integrity, and analysis) and then rates them globally as strong, moderate, or weak per the instrument—articles with no weak domains are categorized as strong, while those with one weak domain are moderate, and those with two or more weak domains are weak overall. After all articles were assessed by two authors, authors RH and KC reached consensus to determine the final scores, and the strength rating was determined by the instruments.

The 11 quantitative articles were assessed for potential meta-analysis. However, there was little overlap in constructs measured. No construct was found across all studies, one construct was found in three studies, and five constructs were found in two studies. Therefore, no attempt was made to combine effect sizes across articles.

Regarding any clinical decisions that were made in the quantitative studies and any reports of decision-making in the more qualitative studies, the authors discussed whether those decisions would be to the benefit (in favor) or to the detriment (against) of the potential patient with mental illness referenced in the studies, and a consensus was reached. No attempt was made to discern the intent of the decision (e.g., whether a patient was referred to a specialist because they needed more specialized care or because the clinician did not want to deal with them), and the authors gave the decision-makers in the study the benefit of the doubt as to their intent in all cases.

Results

Included articles were divided into two groups for synthesis, based upon methods and reported results. The two groups are discussed separately below.

Hypothetical patient studies

In 11 articles, participants responded to a hypothetical patient (presented via a written vignette,^35–41 video,^42,43 or in-person simulated patient)⁴⁴ that did or did not have a mental health diagnosis. Two of these^42,43 utilized the same data but reported on different aspects of the results. Basic details and the most pertinent behavioral results in this group of studies are in Table 1. Articles are listed by global quality and then alphabetically. The final column summarizes whether each behavioral result represented a bias that would favor the medical interests of persons with mental health diagnoses, go against their medical interests, or had neutral implications. There were roughly equal numbers of neutral (n = 17) or detrimentally biased (n = 16) results, but also some favorably biased behaviors (n = 9). Mixed-valence clinical behavior was found within studies. Considering quality ratings, more findings of neutral or positive biases were demonstrated in stronger studies. Weaker studies tended to demonstrate biases that are more negative.

Table 1.

Results of simulation- and vignette-based studies

Author, year	Country, setting	Global quality rating	Relevant participants	Other participants	Hypothetical patient details	Relevant behavioral results	Bias
Crapanzano et al., 201844	USA, Internal Medicine residency programs in California and Louisiana	STRONG	185 Internal Medicine resident physicians	n/a	live simulated new patient (50-year-old female) encounter for chest pain, either with OR without documented history of depression	1) Residents scored lower on OSCE history taking when a history of depression was present	against
						2) Residents scored higher on OSCE patient-physician interaction when a history of depression was present	in favor
						3) Residents correctly diagnosed chest pain in patients regardless of depression history	neutral
						4) Residents created the same work-up plan for patients regardless of depression history	neutral
						5) Residents displayed more disgusted facial expressions when a history of depression was present	against
						6) Residents displayed more neutral facial expressions when there was no history of depression present	neutral
Dixon et al., 200836	UK, University medical school	STRONG	1,239 medical students	n/a	description of a new patient (30-year-old housewife with a 5-year-old child) with history of: well-controlled schizophrenia OR well-controlled depression OR well-controlled diabetes OR no significant medical history	1) Medical students were more likely to agree they would refer the patient to a specialist if a history of schizophrenia was present	in favor
						2) Medical students were less likely to agree they would advise the patient to eat more healthfully and exercise if a history of schizophrenia was present	against
						3) Medical students were more likely to agree they would personally contact the previous GP if a history of schizophrenia was present	in favor
						4) Medical students were more likely to agree they would consider referral for counseling if a history of schizophrenia or depression was present	in favor
Lawrie et al., 199638	Scotland, Randomly selected practicing physicians (mail survey)	MODERATE	52 Primary Care physicians	n/a	description of a new patient (25-year-old single male) with a physical injury from a motor vehicle accident, with subsequent financial troubles, sleep disturbance, tobacco use, AND well-controlled schizophrenia OR no history of mental illness	1) Physicians were more likely to agree they would refer the patient to a specialist if a history of schizophrenia was present	in favor
						2) Physicians were equally likely to agree they would advise the patient to quit smoking regardless of the presence of schizophrenia	neutral
						3) Physicians were more likely to agree they would advise the patient to eat more healthfully and exercise if a history of schizophrenia was present	in favor
						4) Physicians were equally likely to agree they would personally contact the previous GP regardless of the presence of schizophrenia	neutral
						5) Physicians were equally likely to agree they would ask more questions about depression regardless of the presence of schizophrenia	neutral
						6) Physicians were equally likely to agree they would prescribe antidepressants regardless of the presence of schizophrenia	neutral
						7) Physicians were equally likely to agree they would refer for counseling regardless of the presence of schizophrenia	neutral
McDonald et al., 200340	USA, not reported	MODERATE	60 registered Medical-Surgical nurses	n/a	description of a patient (57-year-old male) admitted from the Emergency Department to rule out myocardial infarction with no psychiatric prescription on file OR a list of antipsychotic prescriptions on file OR an antianxiety prescription on file	1) Nurses estimated they would spend an equal amount of time on 14 nursing tasks regardless of the presence of a prescription for an antipsychotic or antianxiety medication	neutral
						2) Nurse reported that they would delegate tasks to ancillary staff equally often regardless of the presence of a prescription for antipsychotic or antianxiety medication	neutral
						3) Nurses were less likely to suspect that myocardial infarction was the correct diagnosis when a prescription for an antipsychotic medication was present than when no prescription or a prescription for an antianxiety medication was present	against
						4) Nurses were most likely to interpret a sudden increase in anxiety as indicative of myocardial infarction when no prescription was present and less likely to interpret increased anxiety this way when a prescription for an antipsychotic or antianxiety medication was present	against
Trachtenberg et al., 201442	USA, Primary Care practices in Massachusetts	MODERATE	256 Primary Care physicians	n/a	video of a new patient (35-or 55-years-old, black or white, male or female) with Type 2 diabetes AND schizophrenia with bizarre affect OR schizophrenia with normal affect OR depression OR eczema	1) When patient and physician characteristics that were balanced in the study design were controlled for, physicians fell roughly evenly into minimalist, middle of the road, and interventionist approaches to Type 2 diabetes care	neutral
						2) When approach type was allowed to vary by comorbidity, physicians who saw a depressed patient more often had an interventionist approach to Type 2 diabetes care	in favor
						3) When approach type was allowed to vary by comorbidity, physicians who saw patients with schizophrenia with bizarre affect more often had a minimalist approach to Type 2 diabetes care	against
						4) When approach type was allowed to vary by comorbidity, physicians seeing patients with either type of schizophrenia never had an information seeking approach to Type 2 diabetes care	against
						5) Physicians who rated schizophrenia as more important than diabetes more often had a minimalist approach to Type 2 diabetes care and none were “interventionists” or “information seekers”	against
Welch et al., 201543						1) Negative attitudes did not predict physicians’ individual clinical actions after controlling for “other patient, provider, and organizational characteristic” (control variables not specified) with the exception of Social Connection (see #2)	neutral
						2) Physicians’ lower feelings of Social Connection to patients was related to a lower likelihood of neuropathy screening	against
Chow et al., 200735	Hong Kong, University teaching/regional acute care hospital	WEAK	322 nurses, 107 physicians	4 unknown occupation	description of a patient (24-year-old male) admitted with headache AND psychiatric symptoms OR diabetes	1) Nurses and physicians agreed they would put equal effort into caring for the patient, regardless of the presence of psychiatric symptoms	neutral
						2) Nurses and physicians agreed they would spend equal amount of time with the patient, regardless of the presence of psychiatric symptoms	neutral
						3) Nurses and physicians agreed they would explain treatment or testing to the patient more carefully when psychiatric symptoms were present	in favor
						4) Nurses and physicians agreed they would explain ward routines to the patient more carefully if psychiatric symptoms were present	in favor
Corrigan et al., 201445	USA, mental health and primary care clinics in 5 Veteran’s Affairs hospitals in the Southeast and Southwest	WEAK	70 Primary Care nurses and physicians	96 Mental Health physicians and nurses	description of a patient (34-year-old male) with hypertension, obesity, disturbed sleep, chronic low back pain attributed to arthritis AND schizophrenia	1) Physicians and nurses who believed a person with mental illness would adhere to treatment were more likely to agree they would refer the patient to a specialist	against
						2) Physicians and nurses who believed a person with mental illness would adhere to treatment were more likely to agree they would refill a pain medication prescription	against
Graber et al., 200037	USA, systematically selected practicing physicians in Iowa (mail survey)	WEAK	232 Family Medicine physicians	n/a	descriptions of a) a patient with severe headache (43-year-old, female) AND no past medical history OR major depressive disorder OR unexplained somatic symptoms; and b) a patient with abdominal pain (62-year-old male) AND no past medical history OR major depressive disorder OR unexplained somatic symptoms	1) Physicians were less likely to suspect a serious illness was present if the patient had a history of depression or unexplained somatic symptoms	against
						2) Physicians were significantly less likely to order further testing for a severe headache if a history of depression was present	against
						3) Physicians were significantly less likely to order further testing for abdominal pain if a history of depression or unexplained somatic complaints was present	against
Macklin & Morrison, 201139	UK, web survey of General Practitioners in 2 cities	WEAK	56 General Practitioner physicians	n/a	descriptions of a new patient with 20% 10-year CVD risk who was a) 44-year-old female with Type 1 OR Type 2 diabetes; b) 48-year-old male with schizophrenia OR epilepsy; and c) 54-year-old male who was retired OR unemployed	1) Physicians recommended statin medications 88% of the time for patients with Type 1 diabetes, 85% for patients with Type 2 diabetes, 37% for patients with schizophrenia, 31% for patients with epilepsy, 33% for retired, and 23% for unemployed	against
Sullivan et al., 201541	USA, mental health and primary care clinics in 5 Veteran’s Affairs hospitals in the Southeast and Southwest	WEAK	146 Primary Care nurses and physicians	129 Mental Health nurses and physicians	description of a patient (34-year-old male) with hypertension, obesity, chronic low back pain, x-ray evidence of arthritis AND schizophrenia OR no history of mental illness	1) Nurses and physicians were less likely to agree they would refer the patient to a weight reduction program if a history of schizophrenia was present	against
						2) Nurses and physicians were equally likely to agree they would refer the patient to a pain management program if a history of schizophrenia was present	neutral
						3) Nurses and physicians were equally likely to agree they would refer the patient for a sleep study if a history of schizophrenia was present	neutral

There was significant diversity and few direct overlaps in behaviors reported across studies. Dixon et al.³⁶ and Lawrie et al.³⁸ had the largest number of directly overlapping behaviors. Both studies used vignettes about patients with schizophrenia, but participants were medical students in one³⁶ and practicing physicians in the other.³⁸ Both articles found that participants were more likely to refer patients with schizophrenia to a specialist, but they disagreed on two other outcomes. Medical students reported a lower likelihood of advising patients with schizophrenia to eat better and exercise,³⁶ while practicing physicians reported a higher likelihood of advising this.³⁸ In contrast, medical students reported a greater likelihood of contacting the prior primary care physician of patients with schizophrenia,³⁶ whereas practicing physicians said they would be equally likely to do so for other patients.³⁸ Lawrie et al.³⁸ had four additional outcomes that were neutrally valenced for people with SMI. Another study showed mixed-valence weight-loss and pain management referral behaviors in a group of nurses and physicians.⁴¹ A third study showed a link between endorsement of stigmatizing characteristics and beliefs about treatment adherence, which was in turn related to provider-reported intent to refer to a specialist and refill pain medication.⁴⁵

Two studies that examined the same data in different ways^42,43 did not find a direct link between negative attitudes and specific clinical decisions to treat diabetes (with one exception, see Table 1).⁴³ Rather, different overall patterns of simulated decision-making were discovered when mental illness was involved.⁴² Specifically, interventionist and information seeking approaches to diabetes care were less common for physicians who read a description of a patient with comorbid schizophrenia. An interventionist approach was more common with a mention of depression. In contrast, another study found that physicians were less likely to suspect serious illness or order follow-up tests for patients who had a history of depression and were described as having a severe headache or abdominal pain.³⁷

Other studies

Of the seven articles summarized in Table 2, six were purely qualitative while the seventh used mixed methods.²⁴ These articles did not inquire about attitudes towards specific mental illnesses (e.g., schizophrenia). Rather, these articles used interviews or focus groups to gather qualitative data on attitudes about mental illness in general. Notably, while Table 1 articles asked participants to make clinical judgments about hypothetical patients, Table 2 articles asked providers to talk about what they observe or actually do in their day-to-day practice.

Table 2.

Results of studies without vignette methodology

Author, year	Country, setting	Quality	Relevant participants	Other participants	Methods	Relevant behavioral results	Bias
Isbell et al., 202031	USA, Emergency Departments	HIGH	42 ED physicians, 30 ED nurses	n/a	Vivid autobiographical memory recall of an encounter with a patient with mental illness; Emotion and engagement measure; Yes/No: “Do you think the emotions that you experienced while treating this patient may have influenced your clinical reasoning and decision-making in this case?”, if Yes or uncertain, describe the influence	1) 9% of physicians and nurses reported acting on their bias while treating a patient with mental illness, but 38% of physicians and nurses reported acting with empathy, patience, and understanding with these patients	in favor AND against
						2) 12% of physicians and nurses reported spending less time with patients with mental illness, but 15% of physicians and nurses reported spending more time with these patients	in favor AND against
						3) 6% of physicians and nurses reported giving unnecessary treatment (e.g., excess sedation) to patients with mental illness	against
						4) 6% of physicians and nurses reported delaying or failing to provide necessary exams or treatment to patients with mental illness, but 27% of physicians and nurses reported advocating for or providing extra care to these patients	in favor AND against
						5) 6% of physicians and nurses reported prematurely closing an encounter for patients with mental illness	against
Lavie-Ajayi et al., 201848	Israel, General Practice clinics, homes, or coffee shops	HIGH	10 General Practice physicians	15 patients with SMI	Interview; question: “In your opinion, in what way, if at all, do severe mental illnesses influence the treatment of physical illness?”	1) Physicians and patients spoke of prejudice against people with SMI that manifested as “downplaying patients’ physical complaints, attributing their health conditions to an expression of their mental health status”; GP quote: “I think that there is [a tendency] to put less effort in preventative medicine... It is easier to think that these complaints are not genuine...”	against
						2) Physicians’ “...notion that patients with SMI find it difficult to take up and utilize available health services” drove “...limited provision of services and referrals.” GP quote: “...they will miss their appointment, then sometimes you even consider making as few referrals and examinations as possible.”	against
Shefer et al., 201449	UK, 4 hospital Emergency Departments	HIGH	19 Emergency Department physicians and nurses	20 Psychiatric Liaison physicians and nurses	Interview; questions: a) “Describe the diagnostic protocol for patients with SMI presenting to the ED”; b) “Does this differ from patients without SMI?”; c) “Describe any cases where a mental disorder interfered with diagnosis of a physical disorder”	1) 13 of 19 (68%) ED staff (physicians and nurses) “...reported one or more incidents in which psychiatric disorder led to misdiagnosis, or delayed examination or treatment with a varied degree of seriousness and with a range of consequences.” Two cases “involved the death of patients who refused to be examined and staff failed to conduct any assessment of their mental capacity to refuse treatment... two other cases of death that staff suspected might have been in consequence of psychiatric disorder interfering with the diagnosis of physical symptoms... also five cases in which delayed diagnosis led to irreversible, long-term damage to the patients’ health.” Eleven “specific cases” of “near misses: were reported in which “the ED staff ‘medically cleared’ a patient and referred him to the psychiatric liaison staff ... whereupon the latter group insisted upon further physical examination during which an organic problem was diagnosed... In other eight cases ... no lasting damage was caused ... but the patient suffered considerable discomfort...”	against
						2) Some psychiatric liaison staff said that they believed that medical staff keep their interactions with patients with SMI to the bare minimum necessary	against
Stumbo et al., 201824	USA, Community Health clinics in 6 states	WEAK quant. HIGH qual	249 Primary Care physicians	163 Primary Care patients	Survey and Interview; goal: “understand attitudes and beliefs about providing preventative care to patients with mental illness	1) Physician quote: “...I probably put preventative care way down on the list because of their mental illness... Because I think that mostly we’re putting out fires about short term things...”	against
						2) Physician quote: “I try to simplify it [prevention message]. So I don’t talk about simple carbs versus complex carbs or the Mediterranean Diet. I usually just say do you drink soda? Don’t drink soda...”	against
						3) Physician quote: “The entire appointment is used to try to counsel the patient on actually taking medication for their psychosis. So then... prevention is often left to the notes in the after visit summary.”	against
van Nieuwenhuizen et al., 201247	UK, Emergency Department	HIGH	8 physicians, 2 nurse practitioners, 15 nurses	n/a	Interview; prompts: a) “Recall times an existing psychiatric illness interfered with the diagnosis of a physical illness”; b) “How does the diagnostic process differ when SMI is and is not involved?”; c) “How can the risk of diagnostic overshadowing be reduced?”	1) Physicians and nurses (64%) felt that diagnostic overshadowing was “a significant phenomenon” but that serious complications as a result of it were rare	against
						2) Physicians and nurses discussed 8 determinants of differential medical care, including “labeling and stigma”	against
						3) Physicians and nurses reported that psychiatric diagnoses and symptoms (including “frequent attender” and “drug or lcohol user” labels) influenced which physical diagnoses were considered or excluded	against
						4) Physicians and nurses linked their own fear of patients with SMI to avoidance behaviors, which also led to patients being less informed about their care	against
Burton et al., 201546	UK, Primary Care clinics	HIGH	16 General Practice physicians, 16 nurses	25 service users with SMI, 11 community mental health staff (CMHT), 7 carers, 6 service user experts	Focus Group; prompt: “Describe experiences reducing cardiovascular disease (CVD) risk in patients with Serious Mental Illness (SMI)	1) “Some primary care health professionals, particularly nurses, described being scared of patients with mental health problems which led to reluctance to offer lifestyle interventions.”	against
						2) “Some service users and CMHT staff discussed experiences of ‘diagnostic overshadowing’ where physical health complaints had been dismissed and attributed to the service user’s mental health problem by primary care health professionals.”	against
Rivera-Segarra et al., 201932	Puerto Rico, inpatient and outpatient metropolitan medical facilities	HIGH	8 physicians, 3 nurses	n/a	Semi-structured interview; specific prompt not provided	1) Physicians and nurses exhibited the theme of attributing physical symptoms to mental illnesses (i.e., diagnostic overshadowing); e.g., “... as soon as they say they are diagnosed with serious mental illness everything they say is like, not true at all.” (physician quote)	against
						2) Related to the theme of diagnostic overshadowing, one physician said “Honestly, we do nothing ... we put them on medication to try to control them so that they let us do the labs and then send them to another place. A psychiatric hospital where they will get the final treatment for that patient.”	against
						3) Related to a theme of lacking skills to work with patients with mental illness, one physician said “...nurses get... so afraid of the patients ... the patient enters the clinic doors and they pass them quickly directly into my office.” A nurse said “Sometimes we put them in an isolated room and restrict them, they are guarded...”	against
						4) Regarding care provided by psychiatrists, one physician said “They don’t pay much attention... they do not check lithium in three years and then when they come back, the patient has diabetes, renal problems, thyroid problems, basically they don’t do follow-ups.”	against

Across the papers included in Table 2, providers reported negative attitudes about people with mental illness. Some physicians and nurses reported a fear of people with S MI^46,47 that patients with mental illness cannot be believed about physical symptoms^31,48 and that patients with mental health concerns exhibit demanding, entitled, and manipulative behaviors in the emergency department.³¹ Providers also reported believing that people with mental illness have trouble accessing medical services, tend to miss medical appointments,⁴⁸ were not interested in preventative care, and would not follow recommendations.²⁴

In several studies, participants linked negative attitudes with differences in clinical decisions in themselves and others. Providers who reported fear reported avoiding patients with mental illness⁴⁷ and reluctance to offer lifestyle interventions and routine health checks.⁴⁶ Providers who distrusted patients with mental illness reported downplaying physical complaints and attributing them to the mental illness.^32,48 Providers who thought patients would have trouble utilizing services reported providing fewer services and referrals.⁴⁸ Providers who believed patients were uninterested in preventative care reported using appointments for the mental health issue and leaving preventative care instructions to the after visit summary.²⁴ Bias against mental illness was felt to be associated with unnecessary treatments such as extra sedation, prematurely closing encounters, delaying or failing to provide necessary exams,³¹ and diagnostic overshadowing by some participants.^32,46–49

Participants reported witnessing severe consequences of attributing physical symptoms to mental illness (i.e., diagnostic overshadowing), including paralysis and death.⁴⁹ Factors that were associated by them with an increased risk of misdiagnosis or delayed treatment in the emergency department included aspects of behavior or appearance of the patient and environmental challenges.⁴⁹ Some providers reported not being aware of important information such as the increase in risk of CVD for persons being treated for SMI.⁴⁶ Reports of diagnostic overshadowing by physicians were often made by others,^46,48,49 not the physicians themselves. However, in the one study where physicians acknowledged it as “a significant phenomenon,” it was also estimated that resultant serious complications were rare.⁴⁷

Discussion

Multiple studies, across clinical settings, have found disparities in medical care for people with SMI.^10–12,16 The purpose of this review was to summarize evidence to support a role for providers’ explicit or implicit biases in clinical decision-making. Results across the 18 relevant studies were mixed. The non-experimental, primarily qualitative studies had generally more negative findings. The studies with an experimental design had a combination of negative, neutral, and positive findings. Interestingly, the non-experimental studies asked providers to report on clinical behaviors they observed in their day-to-day work. In contrast, the experimental studies asked providers to make hypothetical clinical decisions in response to simulations of patients via in-person simulation, videos, vignettes, or written descriptions. The results of the non-experimental studies are in line with the oft-cited view that clinician attitudes have a negative effect on clinical decision-making and are a likely contributor to poor medical outcomes and shortened life spans of people with SMI. The mixed nature of the experimental results does not strongly support this view. Of note, the mixed findings in the experimental studies could be explained by the phenomenon of people “behaving better” when they are observed.²¹

Although the more qualitative studies did find negative bias, it was reported by a minority of participants. For example, in one study, only 9% of physicians and nurses reported acting out their biases with patients while 38% reported acting with empathy, more patience, and understanding³¹. Providers suggested that people with SMI frequently presented with issues such as communication barriers^24,49 and missed a ppointments⁴⁶ that can complicate care, suggesting that such issues are intrinsic to the patient and unrelated to bias. However, it is notable that some provider beliefs were in direct conflict with patient perspectives. Some providers believed people with SMI were not interested in preventative care, whereas patients reported great interest.²⁴ Providers also tended to believe people with SMI miss appointments for personal reasons, whereas patients perceived structural barriers to access, including a lack of outreach from providers.⁴⁶ These results suggest that providers may not recognize their own biases to report them accurately.

Regarding the more mixed results found in studies utilizing vignettes, it is likely that observation effects influenced participants to make more deliberate, socially acceptable hypothetical clinical decisions. Clinical decisions are typically made under more rushed, stressful conditions that allow for more intuitive processing, likely allowing for more influence from biases. Thus, these vignette studies may not reflect accurately what happens in the “real world” with all the complexity of each individual encounter.

Future work in this area should aim to systematically examine factors that could lead to differential treatment for people with SMI. Additionally, the frequency and impact (positive or negative) of differential treatment should be explored. Across studies, clues were present that there were knowledge gaps that could affect clinician judgments.⁴⁶ In three studies, clinicians made decisions related to diabetes care in the context of schizophrenia.^39,41,43 Differences in decision-making could have reflected clinicians being less well versed on the increased risks of metabolic disease in patients with schizophrenia and atypical antipsychotics, rather than a bias against people with mental illness. Incorporation of more training for primary care providers on these risks could help mitigate against this consideration. Consideration for the complexities of some people’s behavior that might interfere with care will also have to be factored in (e.g., poor attention or disruptive behaviors). Care must also be taken to examine the reasoning behind biased decision-making that seems to favor individuals with SMI. For example, more frequently referring patients with schizophrenia to a specialist after a motor vehicle accident than a person without a mental illness³⁸ could be in the patient’s best interest. Alternatively, it could be “moving the patient down the road” to someone else’s care.

Results should also be considered in light of the mixed populations of clinicians and illnesses represented. The interview or focus group studies explored generic “mental illness” in the context of vague illness such as “abdominal pain” or “severe headache.” Biases may be more negative for some disorders (e.g., schizophrenia). The life experiences of the clinicians (personal and family history of mental illness, prior patients with mental illness…) were not explored to determine their impact on the clinical decisions. Ambiguous clinical presentations and limited information in some of the studies may have affected advanced clinical reasoning and decision-making. In addition, outcomes (clinical decisions) varied widely, making cross-study comparisons difficult. Finally, other patient characteristics that could affect decision-making were not controlled (e.g., age and race).

Sampling is another consideration. Study participants ranged from medical students, to practicing physicians of different specialties, to nurses. Level of training is almost certainly relevant. Medical students likely differ from practicing physicians, and all studies should report results for different groups separately, rather than combined as in Sullivan et al.⁴¹

Limitations

There were several limitations to this review. The search strategy was exhaustive, yet there may have been studies not included in the results. This problem may be attributable to differences in terms used to denote stigma against mental illness as opposed to a uniform set of terms across the literature. The search strategy did not include gray literature that may have contained important perspectives.

Implications for Behavioral Health.

Healthcare provider bias against individuals with mental illness does exist and does negatively impact clinical decisions at least some of the time; however, the scale of the problem is unknown, and there may be some positive effects of bias. Further empirical work is needed to definitively quantify how often and to what degree of severity biased clinical decisions impact the lives of individuals with mental illness as well as the impact of personal characteristics or experiences of providers in making clinical decisions. This work should include evaluation of physician decisions in vivo, accounting for the variety of variables influencing those decisions. Future research should also focus on interventions to lessen the influence of provider bias on clinical decisions. Even if negative outcomes are rare, they should be eradicated wherever possible.

Appendix A –. PubMed Search

1. “mental disorders”[All Fields]

2. “mental disorder”[All Fields]

3. “mental illnesses”[All Fields]

4. “mental illness”[All Fields]

5. “psychiatric”[All Fields]

6. “psychological”[All Fields]

7. “mentally ill”[All Fields]

8. “mental health”[All Fields]

9. “psychotic disorders”[All Fields]

10. “psychotic disorder”[All Fields]

11. “schizophrenia”[MeSH Terms] OR “schizophrenia”[All Fields]

12. “schizoaffective” [All Fields]

13. “psychosis”[All Fields]

14. “psychoses”[All Fields]

15. “bipolar disorder”[All Fields]

16. “bipolar and related disorders”[All Fields]

17. “mania”[All Fields]

18. “manias”[All Fields]

19. “manic”[All Fields]

20. “bipolar depression”[All Fields]

21. “depression”[All Fields]

22. “depressive disorder”[MeSH Terms]

23. “depressive”[All Fields] AND “disorder”[All Fields]

24. “depressive disorder”[All Fields]

25. “depression”[All Fields]

26. “depression”[MeSH Terms]

27. “depressive disorder, major”[All Fields]

28. “depressive neurosis”[All Fields]

29. “depressive neuroses”[All Fields]

30. “melancholia”[All Fields]

31. “dysthymic disorder”[All Fields]

32. “mood disorder”[All Fields]

33. “mood disorders”[All Fields]

34. “affective disorder”[All Fields]

35. “affective disorders”[All Fields]

36. OR 2 OR 3 OR 4 … OR 35

37. “physical health”[All Fields]

38. “physical disease”[All Fields]

39. “physical diseases”[All Fields]

40. “physical symptom”[All Fields]

41. “physical symptoms”[All Fields]

42. “psychosomatic”[All Fields]

43. “psychogenic”[All Fields]

44. “diabetes mellitus, Type 2”[All Fields]

45. “diabetes mellitus”[All Fields]

46. “type 2 diabetes”[All Fields]

47. “heart diseases”[All Fields].

48. “heart disease”[All Fields]

49. “prehypertension”[All Fields]

50. “hypertension”[All Fields]

51. “heart attack”[All Fields]

52. “heart attacks”[All Fields]

53. “myocardial ischemia”[All Fields]

54. “myocardial infarction”[All Fields]

55. “myocardial infarctions”[All Fields]

56. “coronary artery disease”[All Fields]

57. “coronary occlusion”[All Fields]

58. “neoplasms”[All Fields]

59. “malignant neoplasm”[All Fields]

60. “malignant neoplasms”[All Fields]

61. “cancer”[All Fields]

62. “cancers”[All Fields]

63. “malignancy”[All Fields]

64. “malignancies”[All Fields]

65. 37 OR 38 OR 39 … OR 64

66. “attitude of health personnel”[All Fields]

67. “health personnel attitude”[All Fields]

68. “health personnel attitudes”[All Fields]

69. “provider attitude”[All Fields]

70. “provider attitudes”[All Fields]

71. “providers attitude”[All Fields]

72. “providers attitudes”[All Fields]

73. “physician attitude”[All Fields]

74. “physician attitudes”[All Fields]

75. “physicians attitude”[All Fields]

76. “physicians attitudes”[All Fields]

77. “nurse attitude”[All Fields]

78. “nurse attitudes”[All Fields]

79. “nurses attitude”[All Fields]

80. “nurses attitudes”[All Fields]

81. “provider perception”[All Fields]

82. “provider perceptions”[All Fields]

83. “providers perception”[All Fields]

84. “providers perceptions”[All Fields]

85. “physician perception”[All Fields]

86. “physician perceptions”[All Fields]

87. “physicians perception”[All Fields]

88. “physicians perceptions”[All Fields]

89. “nurse perception”[All Fields]

90. “nurse perceptions”[All Fields]

91. “nurses perception”[All Fields]

92. “nurses perceptions”[All Fields]

93. “provider bias”[All Fields]

94. “provider biases”[All Fields]

95. “physician bias”[All Fields]

96. “physician biases”[All Fields]

97. “prejudice”[All Fields] OR “implicit attitude”[All Fields]

98. “implicit attitudes”[All Fields]

99. “explicit attitude”[All Fields]

100. “explicit attitudes”[All Fields]

101. “attentional bias”[All Fields]

102. “attentional biases”[All Fields]

103. “cognitive bias”[All Fields]

104. “cognitive biases”[All Fields]

105. 104. 66 OR 67 OR 68 … OR 104

106. (“1990/01/01”[PDAT]: “3000/12/31”[PDAT]) AND “humans”[MeSH Terms] AND English[lang]

107. 36 AND 65 AND 104 AND 105

Appendix B –. PsychINFO Search

1. “mental disorders”

2. “mental disorder”

3. “mental illnesses”

4. “mental illness”

5. “psychiatric”

6. “psychological”

7. “mentally ill”

8. “mental health”

9. “psychotic disorders”

10. “psychotic disorder”

11. schizophrenia

12. schizoaffective

13. “psychosis”

14. “psychoses”

15. “bipolar disorder”

16. “bipolar and related disorders”

17. “mania”

18. “manias”

19. “manic”

20. “bipolar depression”

21. “depression”

22. depression

23. “depressive disorder, major”

24. “depressive neurosis”

25. “depressive neuroses”

26. “melancholia”

27. “dysthymic disorder”

28. “mood disorder”

29. “mood disorders”

30. “affective disorder”

31. “affective disorders”)

32. 32. 1 OR 2 OR 3 … OR 31

33. “physical health”

34. “physical disease”

35. “physical diseases”

36. “physical symptom”

37. “physical symptoms”

38. “psychosomatic”

39. “psychogenic”

40. “diabetes mellitus, Type 2”

41. “diabetes mellitus”

42. “type 2 diabetes”

43. “heart diseases”

44. “heart disease”

45. “prehypertension”

46. “hypertension”

47. “heart attack”

48. “heart attacks”

49. “myocardial ischemia”

50. “myocardial infarction”

51. “myocardial infarctions”

52. “coronary artery disease”

53. “coronary occlusion”

54. “neoplasms”

55. “malignant neoplasm”

56. “malignant neoplasms”

57. “cancer”

58. “cancers”

59. “malignancy”

60. “malignancies”

61. 61. 33 OR 34 OR 35 … OR 60

62. “attitude of health personnel”

63. “health personnel attitude”

64. “health personnel attitudes”

65. “provider attitude”

66. “provider attitudes”

67. “providers attitude”.

68. “providers attitudes”

69. “physician attitude”

70. “physician attitudes”

71. “physicians attitude”

72. “physicians attitudes”

73. “nurse attitude”

74. “nurse attitudes”

75. “nurses attitude”

76. “nurses attitudes”

77. “provider perception”

78. “provider perceptions”

79. “providers perception”

80. “providers perceptions”

81. “physician perception”

82. “physician perceptions”

83. “physicians perception”

84. “physicians perceptions”

85. “nurse perception”

86. “nurse perceptions”

87. “nurses perception”

88. “nurses perceptions”

89. “provider bias”

90. “provider biases”

91. “providers bias”

92. “providers biases”

93. “physician bias”

94. “physician biases”

95. “physicians bias”

96. “physicians biases”

97. “prejudice”

98. “implicit attitude”

99. “implicit attitudes”

100. “explicit attitude”

101. “explicit attitudes”

102. “attentional bias”

103. “attentional biases”

104. “cognitive bias”

105. 105.“cognitive biases”

106. 106. 62 OR 63 OR 64 OR … 105

107. 107. 32 AND 61 AND 106