Dehumanization is the process of depriving someone of positive human qualities. It is usually seen as an overt, active, extreme, conflict-driven phenomenon1 and not typically thought of as being associated with healthcare.

There is however a more subtle form of dehumanization called infrahumanization that can occur outside conflict situations. First described by Leyens,2 infrahumanization differs from blatant dehumanization in that the latter involves the denial of primary emotions (fear, sadness, anger, joy, and disgust) while infrahumanization involves the denial of uniquely human secondary emotions (compassion, admiration, fondness, contempt, and disillusionment).1 Individuals tend to attribute more positive secondary emotions (e.g., compassion) to ‘in groups’ and more negative secondary emotions (e.g., contempt) to ‘out groups’, suggesting that infrahumanization is not simply an expression of dislike but rather a general diminution of the humanity of those perceived to be outside one’s group.2 Infrahumanization can occur whenever groups identify others as being outside their cohort and has been shown to occur in medical settings.3,4 Capozza found that physicians and nurses caring for oncology patients perceived their own professional group as possessing more uniquely human traits than the patient group.3 A study of nursing professionals working in different specialties in an urban hospital showed that denial of uniquely human traits (reasoning, rationality, morality, and intellectual ability) to patients was associated with lower levels of stress among nurses who exhibited high levels of organizational and affective commitment, suggesting that infrahumanization is a subconscious coping strategy to reduce the emotional burden of caring for patients.4

Those suffering from critical illness are particularly vulnerable to this unintentional dehumanization. Several factors contribute:

Critical illness often impairs agency and capability. Loss of agency can be seen as loss of competence and so, although viewed with warmth, those with life-threatening illness may be subconsciously regarded by others as being less human.5,6

Evidence shows that people are more likely to dehumanize those who appear different from themselves.6,7 The physical changes of critical illness, reduced mobility, and a standard hospital gown serve to make patients similar to each other and dissimilar to both their usual selves and their carers. This is further exacerbated by dependency on mechanical devices, breathing tubes, intravenous lines, ventilators, and extracorporeal circuits. Unlike most animals, humans rely on shape and facial recognition rather than smell to identify another human.6 The addition of equipment distorts the basic human shape, may obscure facial features, and impedes our ability to perceive the face as a whole. Piece-by-piece as opposed to whole facial processing is the mode our brain uses to distinguish objects and is known to be a dehumanizing mode of facial perception.8 Functional magnetic resonance imaging has shown that medial prefrontal cortex activation is necessary for social cognition and is activated when seeing other humans but is not activated when seeing objects.6

Critically ill patients are often sedated, reducing their ability to make eye contact. Eye contact is a powerful communication tool that conveys a host of social information. Direct gaze triggers increased activity in the amygdala and activation of approach-related responses, preparing us for social interaction. Averted gaze triggers feelings of disengagement and disinterest.9

Critical illnesses themselves are complex, multifactorial, and difficult to treat. Communication between healthcare professionals requires rapid and accurate appreciation of complicated disease processes. Labelling facilitates this. Mr. Jones, a 52-yr-old farmer with septic shock, diabetes, and renal failure, is often more easily understood as a case of septic shock in a diabetic with acute-on-chronic renal dysfunction.

In an effort to refocus care of the critically ill through a more person-centred lens, several initiatives have been advocated, including diaries, liberal visitation policies, involvement of relatives in basic patient care, and opportunities for relatives to be present during resuscitation.10,11,12 To know if these initiatives allow care providers to see patients in a more human light, we need to determine what effect they have on expressions of empathy among healthcare professionals, empathy being essential to the appreciation of another’s humanity.1,13

In the light of evidence that dehumanization may be protective against burnout and14 that subtle dehumanization may be motivated by the need to reduce the emotional cost of caring,4,15 and recognizing that care of the care giver is a key component of humanized care,12 we also need to find out what effect humanizing interventions have on the risk of burnout and compassion fatigue among both healthcare professionals and relatives.

Finally, as anxiety is common among relatives of the critically ill and can impede critical decision-making,16 and since many of these initiatives focus on increased relative presence and involvement, we also need to determine the effect of humanized care on anxiety among relatives. The objectives of this review were therefore to systematically review the literature to determine the effects of interventions that humanize care of the critically ill on empathy in healthcare professionals, burnout and compassion fatigue among healthcare professionals and relatives, and anxiety among relatives.

Methods

This systematic review was conducted in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses PRISMA guidelines.17,18 A study protocol was designed and was not published separately (available as Electronic Supplementary Material [ESM] appendix 1- Methods section of protocol).

Study type

We included clinical studies (both with and without comparator groups) that measured our pre-specified psychologic outcomes of interest. We excluded non-human studies, studies that did not measure our outcomes of interest, and studies that expressed outcomes qualitatively rather than quantitatively. We applied no language, geographical, or report age restrictions.

Participants

We included studies that measured our outcomes of interest among healthcare personnel who cared for adult (≥ 18 yr of age) critically ill inpatients or among the relatives of these patients. We defined a relative as ‘any person (adult or child), related or otherwise who identified themselves or was identified by the healthcare team as being part of the circle of support for the critically ill person and was described in any eligible study was being a relative or family member’.

Interventions

We defined a humanizing intervention as ‘one with substantial potential to increase physical or emotional proximity between the patient and healthcare professionals or between the patient and their relatives’. We chose this definition on the basis of evidence that increased physical and psychologic distances are key contributors to dehumanization.19,20 Both greater and higher quality inter-group contacts are well supported in the psychologic literature as the most reliable method of overcoming dehumanization.1,21,22,23

While interventions that increase physical proximity (e.g., rounds in the patient’s room, presence of relatives at the bedside, etc.) are easy to identify, those that increase emotional proximity are more subjective. We therefore used a list of potentially humanizing interventions that were agreed on a priori by all authors (Table 1). The list was recognized from the outset as not exhaustive and therefore any non-listed intervention that any author felt might have humanizing potential was discussed with at least one other author. The agreement of at least two authors was required to include or exclude that intervention.

Table 1 Humanizing interventions

We excluded studies of interventions broadly directed at enhancing the emotional or physical health of healthcare professionals or relatives without substantial potential to increase physical or emotional proximity to the patient. We therefore excluded interventions designed to enhance general coping skills or reduce stress among healthcare professionals or relatives including educational programs, resilience training, relaxation techniques, mindfulness training, general communication improvement strategies, stress management programs, and rearrangement of work patterns. Although some of these interventions may indirectly increase appreciation of the patient’s humanity, we felt that it would be impossible to separate out the effects of any indirect humanization from other effects.

Outcomes

To ensure our research question had sound scientific rationale and to optimize the chance that that any association observed was likely to be real, we focused on psychologic outcomes that had either been shown in studies of other populations to be associated with humanization (empathy, burnout, compassion fatigue)13,14,15,24 or were outcomes that had a plausible association and were likely to develop while the patient was still in hospital (anxiety).

Primary

Empathy among healthcare professionals measured by any validated method.

Secondary

Burnout and compassion fatigue among healthcare professionals and relatives, anxiety among relatives, measured by any validated method, and adverse events of any type in healthcare professionals, relatives, or patients. We did not pre-specify adverse events so as to capture all adverse events reported by included studies.

Excluded outcomes

We excluded outcomes that had a plausible but unproven association with humanization and could not be expected to fully manifest while the patient was still in hospital.

Therefore, post-traumatic stress disorder, complicated grief, depression, and family post - intensive care unit syndrome were not addressed by this review.

Although plausibly associated with humanization and likely to occur while the patient was still in hospital, moral distress was excluded because of a well-recognized lack of conceptual clarity and substantial global differences in terminology, making identification of any association prone to significant inaccuracy.25,26

Settings

We included studies conducted in hospital settings in any country. We excluded studies conducted in pre- or post-hospital settings. For hospital settings, we included patients cared for in any critical care setting including medical, surgical, or specialist critical care units. We also included those cared for in postoperative recovery units, recognizing that critically ill patients may sometimes be cared for in this setting (beyond the immediate postoperative period) because of bedspace limitations.

Eligibility

For a study to be included at least two review authors had to agree that it satisfied all of the above criteria for study type, participants, interventions, outcomes, and settings. Studies that did not satisfy all of the above criteria were excluded.

Search strategy

The following databases were searched: MEDLINE, PsycINFO, Embase, CINAHL, Cochrane Central Register of Controlled Trials (CENTRAL), and ProQuest Dissertations and Theses. The initial search covered their dates of inception to 29 June 2016, with an updated search by the same librarian (SMcK) using the same search criteria covering 29 June 2016 to 29 June 2017. Reference lists, conference abstracts, and the World Health Organization (WHO) International Clinical Trials Registry Platform (ICTRP) were searched for additional unpublished studies and ongoing studies (available as ESM; Appendix 2-MEDLINE search strategy).

The full electronic search strategies for all databases can be accessed via QSpace, Queen’s University’s research repository service [http://hdl.handle.net/1974/23768].

Study selection

Search results were uploaded to Covidence systematic review software,27 from where four authors (I.M.G., J.L., R.G., and K.P.) independently screened citations. This was done in two stages. In the first stage, authors independently examined all citations and voted ‘Yes’, ‘No’, or ‘Maybe’ using the Covidence blinded voting system. Citations that received a ‘No’ vote by a minimum of two authors were excluded at this stage. All other citations went on to full-text screening.

In the second stage, two authors (I.M.G. and J.L.) independently examined the full-text version of the studies selected in the first stage to determine eligibility. Full-text articles that received two ‘Yes’ votes were included in the review, those that received two ‘No’ votes were excluded, and those that received any other combination (‘Yes/No’, ‘ Yes/ Maybe’, No/Maybe’, or Maybe/Maybe) were examined by a third author (K.P. or R.G.) to determine eligibility. Any conflicts were resolved by discussion among all four authors. To reduce the risk of missing relevant outcomes that might have been measured but not reported by otherwise eligible studies, one author (J.L.) contacted authors of studies of humanizing interventions that did not report our outcomes of interest.

Data extraction

Two authors (I.M.G. and J.L.) independently extracted data from the included studies using a comprehensive data extraction form. Any conflicts were resolved by discussion among the authors.

Risk of bias assessment

Two authors (I.M.G. and J.L.) independently assessed risk of bias for all included studies using Cochrane’s tool for assessing risk of bias as described in the Cochrane Handbook of Systematic Reviews for Interventions.28 Any discrepancies were resolved by discussion. For each included study, bias was assessed in the following seven domains, random sequence generation, allocation concealment, blinding of participants, blinding of outcome assessors, losses to follow-up, reporting bias, and other bias, and deemed to be ‘low’, ‘unclear’, or ‘high risk’, using a priori criteria based on the effect that bias in each domain may have on the validity of the primary outcome (i.e., empathy).

An overall high risk of bias was defined as either a high-risk rating in four or more of the seven domains of bias or a high-risk rating in at least three domains and an unclear risk rating in one or more domains.

Study quality of individual cohort studies

Additionally, for included cohort studies we assessed quality using the Newcastle Ottawa Scale (NOS), a nine-point scale that assigns scores in three domains: selection, comparability, and outcome.29 A score of 7 or above was considered high quality; 5-7, moderate quality; 4 or less, low quality.

Data synthesis

We planned to calculate pooled estimates of effect for the above outcomes where all of the following conditions are met:

  • Outcomes were reported by at least two studies (both judged to have an overall low or moderate risk of bias) using comparable scales and time points.

  • Absence of substantial clinical or methodologic heterogeneity between included studies.

Clinical heterogeneity was assessed by comparing participants, settings, interventions, outcomes, timing of outcome measures, and ancillary treatments. Methodologic heterogeneity was assessed by comparing risk of bias. Where quantitative analysis was feasible, we planned to assess statistical heterogeneity by visual inspection of forest plots, the Chi2 test, and calculation of the I2 statistic, with a P value < 0.1 in the Chi2 test and an I2 statistic > 50% being indicative of significant statistical heterogeneity. Where significant statistical heterogeneity was present, we planned to present the pooled estimate of effect with subsequent discussion as to the likely impact of heterogeneity on the accuracy and quality of the effect estimate.

Where studies reported outcomes on a continuous scale, we planned to calculate mean differences and standardized mean differences, where studies used the same and different scales of measurement, respectively. Where studies reported the outcomes as dichotomous variables we planned to calculate risk ratios. We planned to present all pooled estimates of effect with their respective P values and 95% confidence intervals. Results were considered statistically significant if a P value < 0.05 was achieved. We planned to perform all meta-analyses using a random effects model in Cochrane statistical software Revman 5.3.30 No subgroup or sensitivity analyses were planned.

Results

Study selection

Search results and study selection are depicted in the study flow diagram (Fig. 1).

Fig. 1
figure 1

Study flow diagram

The initial search returned 11,243 articles and the updated search returned 992 articles, yielding 12,235 articles. Of these, 3,084 were duplicates, leaving 9,151 articles. Nine thousand forty-one were excluded on examination of their abstracts. Articles excluded at this stage included non-human studies, commentaries, letters, and studies in which patient participants were < 18 yr of age.

The remaining 110 articles were assessed for eligibility by examination of their full-text format. Of these, 98 were excluded. Reasons for exclusion were as follows: 41, no intervention tested; 15, outcomes of interest not addressed; three, systematic reviews (these were examined for additional eligible studies and none were identified); one, duplicate study; one, pre-hospital study; 37, intervention did not have substantial potential to increase physical or emotional proximity to the patient. Details of excluded interventional studies are provided (Table 2). 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78,79,80,81 The remaining 12 studies were included in the review.82,83,84,85,86,87,88,89,90,91,92,93

Table 2 Excluded interventional studies

Summary of included studies

The characteristics of included studies are described (Table 3) with further details provided

Table 3 Characteristics of included studies

(available as ESM; Appendix 3). Only two were randomized-controlled studies,82,83 five were pre- and post-intervention cohort studies,84,85,86,87,88 four were prospective cohort studies with non-randomized controls,89,90,91,92 and one was a retrospective study.93 Four discrete humanizing interventions were assessed: diaries,83,84 liberalization of visitation,87,88,89 witnessed resuscitation,82,92,93 and family participation in basic care.85,90,91 One study of a mixed intervention with both humanizing and non-humanizing initiatives was included on the basis that some of the interventions tested had potential to increase physical and emotional proximity to the patient (liberal visitation and family involvement in basic care).86 Ten studies measured anxiety among a total of 1,055 family members.82,83,84,85,88,89,90,91,92,93 Two studies measured burnout among a total of 288 critical care professionals86,87 No studies addressed empathy or compassion fatigue. Two studies of witnessed resuscitation sought unspecified adverse psychologic effects among relatives and reported none.82,93

Risk of bias in individual studies

Eleven of the 12 included studies had an overall high risk of bias.83,84,85,86,87,88,89,90,91,92,93 Details of risk of bias assessment for each study are provided (Table 3) (Fig. 2). Commonly occurring themes were lack of randomization, lack of valid control groups, unclear outcome priorities, and failure to provide sample size justification.

Fig. 2
figure 2

Risk of bias in included studies.

Study quality in individual cohort studies

For the ten included cohort studies, none achieved a high-quality rating; five studies received a score of 5,84,85,87,90,93 putting them at the lower end of a moderate rating, and five had a low quality rating86,88,89,91,92 (Table 4).

Table 4 Risk of bias in individual studies

Outcomes

No eligible studies were found that measured empathy or compassion fatigue in healthcare professionals or relatives.

Burnout among healthcare professionals and relatives

Two studies addressed burnout among healthcare professionals; none addressed burnout among relatives.86,87 Both studies used the Maslach-Jackson Burnout Inventory (MBI) to measure burnout.94 Gianni studied liberal visitation (a minimum of eight hours a day).87 Locally organized staff training sessions were provided at each centre before the change. Burnout among healthcare professionals was measured before and at six and 12 months after the policy change. Baseline levels of burnout were higher among nurses than physicians. A small but significant increase in burnout levels was seen in the year following the policy change, with the increase being greater for nurses than physicians. Of note, staff perceptions of liberal visitation reflected burnout levels with those with high burnout scores expressing more negative opinions than those with low burnout scores.

Quenot also used a before-and-after cohort study to examine the effects of an ‘intensive communication strategy’ that included unrestricted visiting hours, greater family involvement in basic patient care, educational sessions, more frequent family meetings, staff debriefing, role playing, and working groups.86 The strategy was designed to improve communication among healthcare professionals, patients, and families and was designed in response to information obtained by psychologists, who interviewed staff beforehand to identify specific areas for intervention and who remained available for consultation during the study period. The prevalence of severe burnout and depression among staff decreased significantly after the intervention, with all three components of the MBI (emotional exhaustion, depersonalization, and personal accomplishment) showing a significant change. As both studies had an overall high risk of bias and investigated interventions with important differences in their content, a pooled estimate of effect was not calculated.

Anxiety among relatives of the critically ill

Ten studies addressed this outcome.82,83,84,85,88,89,90,91,92,93

Of these ten studies, two measured the effects of liberal visitation on anxiety among relatives,88,89 using the State-Trait Anxiety Inventory.95 Only one reported the actual figures for this outcome and the timing of outcome measurement.89 One study found a significant decrease89 and the other a non-significant decrease in anxiety.88

Two studies evaluated the effects of family-maintained diaries on anxiety among relatives.83,84 One measured the prevalence of anxiety at three months after patient discharge from critical care, using a score of > 8 on a self- administered questionnaire as an indicator of anxiety.84 The other measured mean anxiety levels on the third postoperative day, using the State-Trait Anxiety Inventory.83 Both found a non-significant decrease in anxiety in the diary groups.

Three studies measured the effects of family involvement in basic patient care on anxiety among relatives.85,90,91 Two used the State-Trait Anxiety Inventory,85,90 and one used the Hospital Anxiety and Depression Scale.91 Timing of outcome measurement was reported by only one study.91 All reported significant results.

Three studies investigated the effects of family presence during resuscitation on anxiety among relatives.82,92,93 Two used the State-Trait Anxiety Inventory,92,93 and one used the Hospital Anxiety and Depression Scale and Beck Anxiety Inventory.82 Timing of outcome measures ranged from 48 hr to nine months. None reported significant results.

No pooled estimates of effect were calculated for this outcome as no two studies (each with less than a high risk of bias) of any of the included interventions measured the same outcome using comparable scales at comparable time points.

Adverse events

No adverse events were reported by the two studies that sought them; hence, no pooled estimate of effect was calculated.82,93

Overall quality of evidence

The overall quality of the available evidence was poor with evidence being either low or very low quality. Details and reasons why the evidence was downgraded are provided (summary of findings in Table 5).

Table 5 Study quality – Newcastle- Ottawa quality assessment scale for cohort studies

Discussion

Summary of main results

We found insufficient evidence to make any quantitative assessment of the effect of any of the included humanizing interventions on empathy among healthcare professionals, anxiety among relatives, or risk of burnout and compassion fatigue in either group. We observed a trend towards reduced anxiety among family members who participated in basic patient care, liberal visitation, and diary keeping. Nevertheless, the quality of the data was low and although the trend was consistent it was not consistently significant. We found conflicting effects of liberal visitation on burnout among healthcare professionals again with the limitation of low data quantity and quality.

Agreements and disagreements with other studies or reviews

To our knowledge, this is the first systematic review that aimed to measure the effects of a range of interventions with potential to humanize care of the critically ill on the psychologic wellbeing of healthcare professionals and relatives. This limits the scope for comparison with similar studies. There are however studies that have addressed similar but more focused aspects of this question. Considering these in terms of individual interventions:

Diaries

Ullman and colleagues reviewed the effect of diaries on critically ill patients and their relatives.96 They found only three eligible studies, only one of which addressed psychologic outcomes in relatives. This single study showed a reduced risk of post-traumatic stress in relatives of patients in the diary group.97 They found no eligible studies addressing other psychologic outcomes. Their review differs from ours in that they focused on diaries alone and only included randomized-controlled studies. Aiken et al. did not limit inclusion criteria to randomized-controlled studies and found 11 eligible studies examining the effects of diaries on psychologic outcomes.98 The majority of these studies reported open-ended patient perceptions and were not designed to measure specific psychologic outcomes. Our own finding of a non-significant trend towards reduced anxiety among family members of critically ill patients is supported only by the two included studies that addressed this outcome83,84 and should be interpreted in the light of the substantial limitations in both quality and quantity of the contributing data (Table 6).

Table 6 Summary of findings for reported outcomes

Liberal visitation

In their review of the benefits of flexible visitation in intensive care, Errasti-Ibarrondo et al.99 found 15 studies describing positive responses from families including increased satisfaction, reduced stress, and reduced anxiety. Their review aimed to identify themes rather than quantify effects. Flexible visitation appears to be well received by relatives,99,100,101 but the evidence tends to be observational, exploratory, and limited by substantial bias, making it impossible to be sure of its effects in terms of measurable, reliable outcomes. There is a paucity of research regarding the effects of liberal visitation on healthcare professionals. We found only one study addressing the effects of liberal visitation alone on the risk of burnout among medical professionals; this study reported a small but significant increase in risk of burnout when visiting hours were increased to a minimum of eight hours.87 Interestingly, another study of a multifaceted program that included, among other things, more liberal visitation showed a lower prevalence of severe burnout among critical care professionals.86 The potentially important difference was that liberal visitation here was only one aspect of a comprehensive quality improvement program designed in response to issues raised by healthcare professionals, benefited from psychologic expertise from the outset, and included comprehensive staff training and support sessions with a focus on improved communication between healthcare professionals and relatives. This may explain why this study showed a decrease in burnout levels while the other showed an increase. It is both logical and plausible that certain new practices, no matter how desirable, bring with them ‘change stress’ and that unless introduced thoughtfully and with a change support structure may have unanticipated undesirable consequences. It may well be that it is not liberal visitation per se that increases the risk of burnout among healthcare professionals but rather inadequately supported liberal visitation.

Family participation in care

The three studies included in our review showed a consistent trend towards reduced anxiety among family members who had the opportunity to participate in the basic care of their critically ill relative.85,90,91A narrative review of the literature identified an overall theme of increased satisfaction and reduced anxiety among relatives who participated in care. Nevertheless, this review was exploratory and did not attempt to define outcomes or synthesis effects.102

Witnessed resuscitation

The three studies included in our review showed no significant effect of witnessed resuscitation on anxiety among family members.82,92,93 A Canadian Critical Care Society position paper on family presence during resuscitation based on a literature search of the topic found that in general family members were supportive of the opportunity to be present during the resuscitation of their loved ones and physicians and nurses were largely supportive of the practice.11 The summary recommendation was that the practice was ethically sound and should be considered an important component of patient- and family-centred care. The review upon which this recommendation is based differs from our review in that it included randomized-controlled studies only and included family members of patients resuscitated in a pre-hospital setting and family members of pediatric patients.103 The primary outcomes were patient mortality and quality of resuscitation with family psychologic outcomes being secondary.

What this review adds to the current state of knowledge

Our review highlights the current lack of knowledge regarding the effects of humanized care of the critically ill on empathy and compassion fatigue and provides some potentially useful observations regarding its effects on anxiety and burnout. We found a consistent trend towards reduced anxiety among family members who were allowed more time to visit or were involved in diary keeping or basic patient care.83,84,85,88,89,90,91 Although these studies were small and had a high risk of bias, the consistent direction of their results may well imply a true effect. Interestingly, witnessed resuscitation did not reduce anxiety,82,92,93 which makes sense considering the acute emotional turmoil one is likely to experience when watching vigorous attempts to bring a loved one back to life.

The other interesting observation was the opposite directions of effect we found in the two studies that reported burnout among healthcare professionals,86,87 which may speak to the importance of change support strategies when introducing any new initiative no matter how ‘right’ it feels. Taking these results to the bedside of the critically ill person, it’s probably fair to say that involving families and providing comprehensive support for staff who facilitate greater family presence and participation are likely important contributors to humanized care.

Strengths

Novel and important area of research

First, this is a novel and important area of research. Evidence from psychologic studies is often not widely known to clinicians so their relevance to medical practice can be all too easily missed. When we think of translational research we often think of basic science studies, but translation can and should occur in multiple dimensions. Psychology, the study of human behaviour, has particular value as medicine, no matter how technologically advanced, is ultimately delivered by humans. Infrahumanization is well known to psychologists. This study helps to introduce it to medical clinicians and highlight its relevance to critically ill people and those who care for them.

While many studies have investigated various initiatives aimed at making care more person centred, few have identified these interventions as potentially humanizing, investigated a range of such interventions, or addressed quantitative outcome measures that are both clinically relevant and likely to be influenced by humanization.

Quantitative outcomes addressed

Second, although qualitative studies are ideal for exploring, describing, and understanding perceptions and behaviours, they tend to be less generalizable and are not designed to measure specific defined outcomes, magnitude, or directions of effect or to inform risk-benefit estimation.

Examining humanization through a quantitative lens allows important effects that may otherwise be missed to be captured and measured.

For example, some studies that measured quantitative patient outcomes when families were present during resuscitation found concerning results. A study of simulated codes where the relative present was displaying an overt grief reaction showed delayed time to defibrillation and fewer total shocks delivered.104 A retrospective study of more than 300 patients in a medical intensive care unit showed significantly lower rates of return of spontaneous circulation and survival to hospital discharge for patients whose families were present during cardiopulmonary resuscitation.105 Although these results are not consistent across studies, they do raise an important issue. While humanization is innately right, it does not necessarily follow that all efforts to optimize it should be pursued without considering the risk-benefit balance. As for all interventions, we need to know which work best, what the effects are (good and bad), the size and direction of those effects, and for whom the benefits outweigh the risks. Quantitative data about clinically relevant outcomes can provide us with this information.

Robust methodology

Lastly, this review was conducted using a rigorous methodology in accordance with the PRISMA guidelines for systematic reviews.19 The search was not restricted on the basis of language, study date, or location, ensuring maximum retrieval of studies that met our inclusion criteria.

The use of a blinded voting system to select studies minimized the risk of selection bias and the use evidenced-based criteria to define humanizing interventions allowed precise appraisal of the effects of interventions that were likely to have substantial potential in this regard.

Limitations and potential biases

This review has a number of limitations and potential biases

Outcome-based search strategy

Our search strategy was population and outcome rather than population and intervention focused. We chose this approach for two reasons. First, we felt that it allowed us to identify all interventions with humanizing potential that evaluated our outcomes of interest and reduced the risk of missing studies that evaluated interventions that we had not thought of as humanizing a priori. Second, it allowed us to identify a manageable number of studies to screen. Searching for the concept of ‘humanization’ using terms such as humaniz* or respect* retrieved over one third of OVID Medline’s records, which is approximately 9 million citations (where * allows for any truncation of the word). While our choice of search strategy was both logical and feasible, we cannot exclude the possibility that we missed studies that measured our outcomes of interest but described them without using the specific labels of empathy, burnout, compassion fatigue, and anxiety. To mitigate this risk, for studies examined in full-text format and excluded on the basis that they did not report our outcome of interest, one author contacted study authors to see if they measured our outcomes of interest but expressed them in other ways. None were identified.

Focused definition of humanizing interventions

To get a clear view of the effect humanizing interventions on healthcare professionals and relatives, it was necessary to select interventions that had substantial potential to increase the awareness of others of the patient’s humanity. For this reason, we only included studies that had substantial potential to increase patient-healthcare professional or patient-relative inter-group contact through increased physical or emotional proximity. We excluded studies of interventions that were broadly directed at enhancing the emotional or physical health of healthcare professionals or relatives without substantial potential to increase physical or emotional proximity to the patient. This meant excluding educational programs, resilience training, relaxation techniques, mindfulness training, general communication improvement strategies, stress management programs, and rearrangement of work patterns.

Defining humanizing interventions in this way meant that any humanizing potential of broader, more care provider- and relative-directed interventions is not captured in this review. While this may be a weakness in terms of measuring the effect of all interventions that may have any humanizing potential, it allowed a more precise appraisal of those with substantial potential in this regard.

Limited outcomes addressed

Humanization is a psychological concept and cannot be fully understood by quantitative methods alone, with both quantitative and qualitative research having important roles to play in furthering our understanding of this area. For this review, we chose to focus on selected quantitative outcomes, choosing to sacrifice breath of knowledge to get a more focused and accurate view of the current state of knowledge in terms of definable, measurable, and reliable outcomes. This meant that research describing the feelings of patients, relatives, and healthcare professionals in descriptive or narrative terms was not captured in this review. We also excluded outcomes that had a plausible association with humanization but were not likely to fully manifest while the patient was still in hospital; we cannot say whether humanization influences late-occurring psychologic outcomes including complicated grief, depression, and post-traumatic stress. Identifying any association between humanization and more long-term psychologic outcomes quantitatively is subject to substantial distortion by confounding factors. Even if known confounders are adjusted for, there are likely to be many unknown confounders that may blur any real association. Qualitative methods may be much more appropriate for investigating those outcomes.

Moral distress was not evaluated by this review and, although there is no proven association, when we consider that moral distress arises in situations where one is required to act in a manner that conflicts with their own belief of what is morally correct, it is plausible that humanized patient care may lessen or exacerbate moral distress depending on the individual caregiver’s perception. This would be worth investigating in future studies.

Conclusions

This review identifies a knowledge gap regarding the effects of humanizing interventions on clinically important quantitative psychologic outcomes among healthcare professionals and relatives. It does however suggest that relatives of the critically ill are less anxious when they feel more involved or are allowed more time at the bedside and that liberal visitation may have opposing effects on healthcare professional burnout dependant on the context in which it is introduced. One should be cautious, however, about basing practice change on the findings of a few, small, generally poor-quality studies. As we move towards increasing patient and family-focused care, it is important we consider the emotional impact on healthcare professionals and families. While the natural human assumption may be that only good things come of person-focused and family-inclusive care, we do not know to what extent that is true. To get an accurate picture of the broader effects of humanized care, we need to study its effects on patients, families, and healthcare professionals both qualitatively and quantitatively.

Differences between the protocol and the review

In response to expert peer review, we revised our original outcomes to include burnout and compassion fatigue among relatives of the critically ill. As our original search used all the terms needed to identify these studies, we did not repeat the search. Two authors rescreened the studies to identify any that might have measured these outcomes among relatives of the critically ill. None were identified.