Respecting the right to refuse: is decision-making capacity disproportionately challenged in patients declining medical care in order to treat?

Abstract

Objective

This study aimed to determine the proportion of capacity challenges for refusal of care, the risk factors that may trigger challenges to capacity and to consider the implications of challenging capacity.

Methods

This retrospective cohort study examined all psychiatric consults placed for decision-making capacity assessment over a 2-year period in an urban community hospital. Data were collected on the reasons for consultation, patient demographics and whether treatment occurred. Statistical analysis, including χ² test for equal proportions, was done to determine whether the proportion of capacity challenges for refusal of care differed from 50%.

Results

122 consults questioning capacity were identified during the 2-year period, of which 112 (91.8%) were placed in the setting of refusal of care. Males were over-represented (68.9%) in the consults placed for capacity assessment compared with the proportion of males in all hospital admissions (48.8%) (p<0.0001). The odds of being treated increased in the cohort of patients who lacked decision-making capacity compared with those who were deemed to have capacity (OR 4.2 (95% CI 1.4 to 10.4); p=0.002).

Conclusion

This study shows that capacity is challenged overwhelmingly in patients declining medical care. When these patients are decentred from decision making, the odds of completing treatment were more than four times higher.

Introduction

Medical decision-making capacity (DMC) is a foundational component of informed consent.¹ The clinical practice of assessing capacity is intended to uphold the ethical principle of autonomy and protect patients from making value-incongruent decisions during a period of possible impairment. To standardise the approach to capacity assessment, the four skills model of decisional capacity has emerged as the predominant assessment model in clinical practice.^{2 3} Although there has been some critique of this model as favouring rationality, its wide acceptance in clinical practice has contributed to its statutory codification in many jurisdictions.^{4 5} Assumed in the current clinical adoption and scholarship, relating to capacity is its empiric value in upholding patient autonomy. Though typically considered a brief and benign clinical interview, emerging literature suggests fundamental risks and dangers associated with the exercise of evaluating decisional capacity.⁶ For example, capacity may be disproportionately challenged in black, Indigenous and people of color, perpetuating structural oppression and racial disparities in healthcare.⁷ Others have suggested that there is virtually no ethical value to evaluating DMC as it stands.⁸

Refusal of recommended medical care alone does not indicate lack of DMC, nor is challenging capacity only necessary in instances of refusal of medical care.⁹ In theory, physicians should request assistance with determining ability to engage in informed consent for cognitively impaired patients regardless of agreement with the medical plan. Therefore, logically, clinical application of capacity challenges should not disproportionately favour instances of refusal.

This assumption reflects an internalist understanding of DMC. In this view, capacity is determined by whether a patient demonstrates the core abilities of understanding, appreciation, reasoning and communication, regardless of whether their ultimate decision is to accept or refuse treatment. Refusal, by itself, does not raise the threshold for DMC, and thus, in theory, one would expect similar proportions of consults following acceptance and refusal of care. An externalist account, on the other hand, holds that the threshold for demonstrating capacity should be higher when a patient’s choice is expected to lead to harm. From an externalist perspective, refusals of medically reasonable treatment are treated as ‘flags’ that trigger capacity assessments, and one would expect a disproportionate number of such referrals. Here, we frame our analysis within the internalist four-abilities model but test whether referral patterns show an externalist bias towards challenging refusals.

While studies have examined the reason for capacity consultation in the past, none to our knowledge have examined the proportion of capacity challenges for refusal versus consent.¹⁰ We hypothesised that DMC evaluations are disproportionately initiated following refusals of recommended care. The process of psychiatric consultation in the medical setting is riddled with disparities.^{11 12} Suspicions that capacity challenges may be used as a means to override decisions that are discordant with a medical provider’s recommendations are not new.¹³ This study aimed to determine the proportion of capacity challenges for refusal of care, the potential risk factors that may trigger challenges to capacity and to consider the ethical and clinical implications of challenging capacity.

Methods

A retrospective chart review examined all 122 psychiatric consults placed to the consultation-liaison psychiatry service at a safety-net community hospital in New York City for assessment of medical DMC from 1 January 2022 through 15 December 2023. Deidentified data were extracted from the electronic medical record (EHR), including patients’ age at time of consult, sex at birth, self-identified gender, self-identified race and ethnicity, language spoken, highest level of education attained, employment status, date of psychiatric consultation and dates of admission during which the psychiatric consultation occurred, primary medical and psychiatric diagnoses, reason for psychiatry consultation, completion or deferment of a capacity assessment, the determination of DMC and whether the patient was treated or not. Patients’ preferred gender was used in analyses unless data were unavailable (18%), in which case sex at birth was substituted. Psychiatric diagnoses were classified into psychotic disorders, mood disorders, substance use disorders, personality disorders and neurocognitive disorders. Medical diagnoses were categorised as neuropsychiatric, trauma or surgery related, infectious or other general medical disorders (eg, respiratory, renal or cardiac). The study was approved by the institutional review board (IRB) of the Biomedical Research Alliance of New York. Because the study was of less than minimal risk, the IRB determined that written informed consent could be waived. Data from this study are available from the corresponding author on reasonable request. The data are not publicly available to protect the privacy of the included patients.

The expected proportions for demographics were obtained using the Slicer-Dicer tool within the EPIC EHR. Aggregate census data was generated using a report query for all admissions between 1 January 2022 and 31 December 2023 to the specific hospital units served by the consultation-liaison psychiatry service and ‘sliced’ by top 10 results for race as recorded within the respective charts. The expected proportions determined from the hospital admissions during this time period were 63.2% black, 5.6% white, 27.7% something else and 3.5% no answer. For ethnicity, the expected proportion of individuals identifying as Hispanic from Slicer Dicer census data was 28.4%. The expected proportion for gender was 48.8% male during this time period.

A post-hoc sample size calculation to assess the power of the available sample size indicated that a minimum of 42 patients in each group provided a power greater than 80% with an alpha equal to 0.05 to detect a medium effect size (difference in proportion of 0.3) using Fisher’s Exact test. Categorical variables were described as counts and proportions; age was summarised as mean with SD, but length of stay, which had a skewed distribution, was summarised as median with IQR. Binomial exact tests or a χ² goodness-of-fit test were used to both determine if the demographic proportions of the study population differed from hospital census data and whether the proportion of capacity consults specifically for refusal of care differed from 50% with the internalist assumption that, if the introduction of the question of capacity was uncoupled from the refusal of medical care, patients would be equally likely to have refused care versus not. Fisher’s Exact tests were used to determine differences between categorical groups. Student’s t-tests or 1-way ANOVA were used to compare differences in age between groups; Kruskal-Wallis tests were used to assess differences in length of stay. For variables with more than two groups, post-hoc pairwise comparisons were done with p values adjusted using the Bonferroni method. An exploratory multivariable logistic regression model was created to assess predictors of lack of decisional capacity from any statistically significant univariable associations and accounting for age, gender, race, ethnicity and reason for the consult questioning capacity. Similarly, multivariable logistic regression models were used to assess predictors of whether patients received recommended care accounting for age, gender, race, ethnicity and presence or absence of decisional capacity. The statistical analyses were performed using R Studio (V.2023.12.0+402) and generated 2-sided p values. The R script is available in the online supplemental material.

Results

Of the 122 consults placed for assessment of decisional capacity, 68.9% of patients identified as black, 8.2% as white, 21.3% as ‘something else’ and 1.6% chose not to answer. These were not substantially different from the expected proportions of self-identified race determined from all hospital admissions during the study period (p=0.19). For ethnicity, 16.8% of patients self-identified as Hispanic, which was lower than the expected proportion (28.4%) of Hispanic patients determined from all hospital admissions (p=0.005). Race and ethnicity were considered separately, as they are collected as discrete categories within the EHR. Males were over-represented (68.9%) in the consults placed for capacity assessment compared with the proportion of males in all hospital admissions (48.8%) (p<0.0001) (online supplemental table).

A majority (91.8%; n=112) of capacity consults were placed for reasons involving refusal of recommended care (figure 1; p<0.0001 for difference from 50%). Examples of recommended care included disposition, lab work, imaging, fasting in preparation for a procedure, nasogastric tube placement, oesophagogastroduodenoscopy, dialysis, N-acetylcysteine after an acetaminophen overdose, surgical treatment of an abscess and limb-saving surgery. The proportion of black, white and ‘something else’ patients for whom such a consult was placed was not statistically different from expected proportions; however, 17.4% of these patients identified as Hispanic, which was once again lower than expected based on all hospital admissions during the study period. Of the patients for whom capacity consults were placed for refusal of care, 69.6% of patients were male compared with the expected 48.8% based on all hospital admissions (p<0.0001). No specific medical or psychiatric diagnoses were found to be associated with consultation requests for capacity assessment in the setting of refusal of care.

Figure 1. The overwhelming majority of consults to determine decisional capacity were initiated because of patients’ refusal of care. P value <0.0001 for test of equal proportions.

Of the 122 capacity consults placed, 28 were deferred and 94 (77.0%) were completed. Consults were deferred at the discretion of the consulting psychiatrist if it appeared that challenging capacity was being done without appropriate disclosure, without ample opportunity for communication or with unrealistic expectations for the outcome of the capacity challenge. Of the 94 consults completed, 51.7% of patients (n=46) were deemed to have decisional capacity while 48.3% (n=43) were assessed to lack it (p=0.75; figure 2). Documentation was inconclusive for five patients. In univariate analyses, a diagnosis of a neurocognitive disorder (OR 9.4 (95% CI 3.4 to 25.9); p<0.0001) was associated with inability to demonstrate decisional capacity; mood disorder (OR 0.25 (95% CI 0.09 to 0.72); p=0.009) was associated with ability to demonstrate decisional capacity. However, in multivariable logistic regression accounting for age, gender, race, ethnicity and reason for consult (capacity to refuse treatment or capacity to consent to treatment), only a diagnosis of neurocognitive disorder remained predictive for lacking decisional capacity (OR 13.7 (95% CI 3.5 to 55.6); p=0.0002). Gender, race, ethnicity, language spoken, education, presence of personality disorder, psychotic disorder and substance use disorder were not associated with the presence or lack of decisional capacity.

Figure 2. In the group with completed consults, patients were equally likely to have or not have decisional capacity. P value =0.75 for test of equal proportions, represented by the centre white line.

In those with a completed capacity challenge, treatment outcome (recommended care occurred or did not occur) was known in all but two instances. Absence or presence of decisional capacity was associated with ultimate treatment outcome. The odds of being treated increased when the patient did not have capacity when compared with those who did (OR 4.2 (95% CI 1.4 to 10.4); p=0.002). This association was unaltered after accounting for age, gender and race in a logistic regression model (OR 3.7 (95% CI 1.4 to 9.4); p=0.007). In a subset analysis of only those patients with a consult for refusal of care (n=87), the association was somewhat stronger (OR 5.2 (95% CI 1.9 to 14.8); p=0.002) even after accounting for age, gender and race. Figure 3 demonstrates the trajectory of care for patients with a consult for refusal of care.

Figure 3. Trajectory of completed consults for determination of decisional capacity based on initial reason for the consult (A) including resultant decision on capacity (B) and whether the patient ultimately received the recommended medical treatment (C). ‘NA’ indicates instances in which resultant decision on capacity (B) and whether the patient ultimately received the recommended medical treatment (C) were not explicitly documented in the record. In patients refusing care, the odds of being treated increased when the patient did not have capacity (OR=5.2 (95% CI 1.9 to 14.8); p=0.002).

Discussion

The primary findings of this study are that an overwhelming majority (91.8%) of capacity consults were placed for refusal of medical care and that the odds of being treated increased significantly in patients determined to lack capacity. Although there was no evidence of racial bias increasing referral for capacity, males were overrepresented, and individuals with Hispanic ethnicity were underrepresented in this cohort. Only about half of the patients with a completed consult lacked decisional capacity, which was strongly associated with the diagnosis of a neurocognitive disorder.

The high proportion of consults for refusal of the medical recommendations brings into question the ultimate purpose of and motivations for challenging decisional capacity as well as the ethical implications of this procedure. If capacity challenges are not primarily motivated by a desire to override refusal, one would expect an equal proportion of consults for acceptance and refusal. As the responsibility of informed consent does not discriminate between acceptance and refusal, the near totality of consults (91.8%) being requested for refusal signals a strong preference for challenging capacity when recommended care is refused. Although it is not explicitly clear what motivates this preference in our data set, prior literature illuminates that DMC determinations are not free from the influence of physician personal values.^{14 15} Therefore, it is plausible that the inclination for challenging capacity in refusal betrays an intolerance for patient refusal by physicians who have a proclivity for treatment and underestimate the harms.¹⁶ This may be further understood in light of the internalist-externalist distinction. Referrals were often triggered by frontline clinicians who appeared to treat refusal itself as a ‘flag’ for concern, consistent with an externalist orientation. However, the psychiatric assessments themselves generally adhered to internalist criteria, in which the threshold for capacity does not change based on whether the choice may lead to harm. This could explain why capacity was found lacking in only 48.3% of cases, despite referrals being disproportionately clustered around refusals. This tension between referral and assessment underscores the ethical stakes of capacity challenges: even when final determinations do not confirm incapacity, the act of challenging capacity disproportionately burdens patients who choose to decline treatment. This act of suspending the patient’s spontaneous ability to choose for themselves, even if intended for a beneficent outcome, may still be experienced as a weaponisation of beneficence that decentres the patient from their decision-making.

The potential harms of decentring patients via capacity challenges are dire and well documented in the literature. These harms include the lack of consensus leading to significant variability in outcomes of assessments, ‘value impregnation’ and its propensity to influence challenge outcomes based on physician values, disproportionate introduction of challenges in the care of marginalised populations undermining health equity, surrogate decision makers introducing personal bias by making choices inconsistent with patient values and the erosion of trust that occurs when autonomy is overridden.^{6 7 15 17–19}

This study demonstrated no association between patients’ spoken language or level of education and the risk of losing capacity, although prior studies have suggested that low education level and language barrier may be risk factors for impaired medical DMC.²⁰ There was no evidence of race being a risk factor for increased placement of capacity consults or outcomes of capacity assessments, and these findings may be attributed to the general mission of the NYC Health+Hospitals system to uphold diversity and inclusion and promote equity in healthcare. However, Hispanic ethnicity was underrepresented. This is reminiscent of a separate study about preferred language and psychiatric consults which found that Spanish-preferring patients were underrepresented among the patients for whom a psychiatric consult was requested.¹¹ Both studies could suggest that Hispanic or Spanish-speaking patients were overlooked or that they were less frequently pathologised, and this warrants further exploration. Patients with a neurocognitive disorder diagnosis were more likely to be found unable to demonstrate capacity, which is consistent with prior studies.²¹

Male patients were more likely to have a psychiatric consult for capacity assessment in general (68.9% compared with the expected 48.8%) as well as in the setting of refusal of care (69.6% compared with the expected 48.8%). Various reasons for this may be considered. Studies have shown that men exhibit higher levels of aggression than women.²² It may be speculated that men may express outward aggression more readily in the context of refusal of treatment or general dissatisfaction with their medical care, prompting an implicit (or explicit) desire in the primary care team to involve a behavioural health specialist. In other words, a request for DMC assessment may be an implicit response to a perceived threat from the patient. Another hypothesis may be that women are more commonly convinced or coerced into treatment or are more agreeable with treatment recommendations in the first place. Multiple studies, however, show that women are more likely to refuse procedures such as cardiac catheterisation and hormone therapy and surgery for breast cancer.^{23 24} Men, on the other hand, were found to be more likely to leave the hospital against medical advice after being diagnosed with COVID-19 infection.²⁵

This study has several limitations, including a small sample size and limited representation of demographic groups such as non-English speakers. Because of the small proportion of consults unrelated to refusal of care, this study lacked the power to detect additional associations between patient factors (demographics, medical and psychiatric diagnoses) and the reasons for capacity evaluation and suspension of autonomous decision-making. Because the sample was taken from a single site, results may lack generalisability, especially in settings with different demographics. Aggregate data from the EPIC EHR SlicerDicer tool are difficult to validate.

Disproportionate intrusion on the decision to decline recommended care in conjunction with over-representation of males in capacity consults reveals significant possibility for concentrated interference on specific individuals and their engagement with healthcare. While we did see isolated racial disparities in referral of Hispanic patients for capacity challenges in this data set, prior studies have demonstrated that race, language and age are risk factors for capacity challenges which evoke concerns about harm for vulnerable populations.^{7 20} When examined through a structurally competent lens, one can recognise DMC challenges as a potent instrument for medical hegemony. Deliberately using capacity challenges to suppress instances of refusal is conceivable as it was the anecdotal experience of early thinkers on capacity and even the responsibility of physicians according to some bioethicists.^{13 26} There is strong evidence to support this in our data: when capacity to refuse care is challenged and the patient is successfully displaced from decision-making, that patient is then five-fold more likely to be treated than patients who remained centred in their decision-making.

Data availability statement

Data are available upon reasonable request.