Hospitalisation outcomes for patients with severe mental illness treated by female vs. male psychiatry residents

Abstract

Objective

Recent literature suggests that female physicians provide higher quality of care compared to their male counterparts across a variety of physical medical conditions. We examine whether a similar phenomenon is observed for psychiatry residents treating hospitalised psychiatric patients.

Methods

We analysed 300 hospitalised patient records from Shalvata Mental Healthcare Centre (Hod Hasharon, Israel). Resident-patient sex matchings were compared.

Results

No significant differences were observed in terms of residents’ age and patients’ age, medical condition and hospitalisation history. Male and female patients treated by female residents presented shorter hospitalisations (58 and 54 days compared to 67 and 66 days, respectively, p < .05), longer time to next hospitalisation (269 and 179 days compared to 179 and 123 days, respectively, p < .01), lower 30-day readmission rate (37% and 35% compared to 10% and 19%, respectively, p < .05), higher levels of family involvement during hospitalisation (2.6 and 2.7 points compared to 2.1 and 1.9 points, respectively, p < .01) and higher chances of obtaining rehabilitation services (39% and 34% vs. 23% and 17%, respectively, p < .05).

Conclusions

Hospitalised patients treated by female psychiatry residents are associated with better hospitalisation outcomes compared to those cared for by male residents.

KEY POINTS

• Both male and female patients treated by female residents presented better hospitalisation outcomes.

• These hospitalisation outcomes include shorter hospitalisation periods, longer time to next hospitalisation, lower 30-day remission rate, significantly higher levels of family involvement and higher chances of obtaining rehabilitation services.

• Further work is needed in order to investigate the sources and reasons for the identified differences.

Keywords:

• Hospitalization outcomes
• psychiatry residents
• case management

Introduction

A growing body of literature has shown better outcomes for patients treated by female physicians. These studies range across a wide variety of physical medical conditions and treatment settings, such as lower mortality and readmission rates for patients admitted to internal medicine wards (Tsugawa et al., 2017), lower mortality rates for patients undergoing common surgical procedures (Wallis et al., 2017) and following acute myocardial infarctions (Greenwood et al., 2018), reduced error rates in diagnostic radiology practice (Alcaide-Leon et al., 2022), higher detection rates of adenoma during colonoscopies (Mehrotra et al., 2018), better management of type 2 diabetes (Berthold et al., 2008), fewer emergency department visits and hospitalisations for patients with a female primary physician (Dahrouge et al., 2016), and even better oral hygiene among patients of female dentists (Takeuchi et al., 2020), to name a few. Several possible explanations to account for these differences were posed, including female physicians’ tendency to follow evidence-based medicine guidelines more than their male colleagues (Baumhäkel et al., 2009; Reid et al., 2010; Tsugawa et al., 2017) and their more patient-centred and direct patient communication and interactions (Ganguli et al., 2020; Krupat et al., 2000; Roter et al., 2002). Unfortunately, these and similar possible explanations currently remain as speculations, as a definitive mechanistic link has yet to be established.

Common to these studies and their posed explanations is the primary focus on physical medical conditions, treatments and outcomes. As such, this plethora of research need not necessarily translate well to the mental healthcare domain given its unique nature. Moreover, to the best of our knowledge, no evidence to support any physician sex differences is currently available regarding the field of mental health.

In this study, we set to examine whether treatment by female psychiatry residents is associated with better patients’ hospitalisation outcomes compared to male residents. We choose to focus on the care provided in closed wards since these patients commonly present severe mental illnesses, which require extensive and timely care. This care is normally provided by quasi-randomly¹ assigned psychiatry residents where, in most cases and especially those with acute mental conditions and/or comorbidity, residents act not only as physicians but also as case managers who assess, plan, implement, coordinate, monitor and evaluate their patients throughout their hospitalisation, and, in some cases, even after their discharge (Frankel et al., 2018; McLaughlin-Davis & Message From the Chair, Task Force on the Practice of Hospital Case Management, 2019). This program, which is very popular across the world, is known as the clinical case management program (Kanter, 1989) and is in the focus of this study.

Materials and methods

Data

For data acquisition, we retrieved records from the Shalvata Mental Healthcare Centre (Hod Hasharon, Israel), which implements the clinical case management program (Kanter, 1989). We randomly² selected, and retrospectively analysed, 300 adult patients who were admitted to the hospital’s closed wards during the year 2021 and who had at least one prior hospitalisation in the last five years. We shall refer to this prior hospitalisation as the index hospitalisation, which will be at the focus of our analysis (as discussed next). We restricted our sample to patients who were hospitalised and managed by a psychiatry resident for at least one week, thus allowing sufficient opportunity for meaningful resident-patient interaction. Patients whose case manager changed during their hospitalisation period and those who suffer from severe cognitive impairments or autism spectrum disorder were also excluded, for the same reason. Scheduled hospitalisations, such as routine check-ups, were also excluded.

Our choice to focus on the index hospitalisation (and not the most recent one) for our analysis is intended at facilitating the examination of the patients’ time to readmission (i.e., time from the last discharge to the next admission). Specifically, considering a patient’s most recent hospitalisation would fail to capture the duration in which the patient was not hospitalised. As discussed in section ‘Measures’, this is one of our central measures in characterising hospitalisation outcomes. Also note that our focus on psychiatry residents dramatically reduces the need to account for possible confounding physician-related factors such as physician experience, authority, role, advanced qualification, etc. Specifically, in Shalvata, as is the case in most clinical case management program implementation (Kanter, 1989), most patients are managed by early career residents who are no more than 2 years into the training program. It is also important to note that patients have little agency over their choice of case managers, thus ensuring a quasi-random assignment of case managers.

Measures

Each hospitalisation was represented by the patient, physician and hospitalisation outcome characteristics. Specifically, patients were characterised by their demographics and key indicators of medical history. Physicians were characterised by their demographics. These measures, provided in detail below, are aligned with prior literature (e.g. Greenwood et al., 2018; Tsugawa et al., 2017; Wallis et al., 2017). Unfortunately, accurately characterising hospitalisation outcomes is much harder as different stakeholders may have different views on what constitutes ‘hospitalisation success’ (Perkins, 2001). For example, the primary interest of many families is receiving the support, information, services and involvement that they need to fulfil their tasks as primary caregivers (Shepherd, 1995). On the other hand, clinicians may be more interested in empowering their patients and allowing them to function properly in society for an extended period of time (Rogers et al., 1997). As such, we devise a set of objective measurable metrics as defined below to capture the varying perspective on hospitalisation success. These objectives combine both ‘standard’ measures of hospitalisation success (hospitalisation duration, time to readmission and 30-day readmission rate) as well as some less orthodox ones (level of family involvement during stay and the acquisition of rehabilitation services). Considering the latter, substantial family involvement during psychiatric hospitalisation was found to be associated with post-discharge clinical outcomes such as medication adherence and continued treatment (Drapalski et al., 2009). Similarly, acquiring rehabilitation services facilitates the integration of the patient back into the community with maximal degree of functional independence and quality of life (Dalton-Locke et al., 2021). As such, these measures are also considered to be a part of ‘hospitalisation success’.

Patient characteristics

Sex, age, primary diagnosis (schizophrenia, psychotic disorders, affective and anxiety disorders, personality disorder, other), age at diagnosis, number of previous hospitalisations, mean time of hospitalisation in the last 5 years, mean time between consecutive hospitalisations in the last 5 years are the patient characteristics.

Physician characteristics

Sex and age are the physician characteristics.

Hospitalisation outcome characteristics

Hospitalisation duration, time to readmission, 30-day readmission (binary, indication whether the patient was readmitted within 30 days of the discharge date), family involvement during stay (measured on a four-point ordinal scale: 1 = no family interaction; 2 = a single meeting with the family; 3 = several meetings with the family; 4 = complete family involvement, which includes at least one additional staff member, e.g., a social worker or a psychologist) and the acquisition of rehabilitation services (binary) are the hospitalisation outcome characteristics.

Analysis

Prior studies have also established that patients’ demographics, and especially their sex, are associated with their mental condition, treatment, remission rates and coping efficiency in general (Salzer et al., 2018) and specifically for common psychiatric conditions such as schizophrenia (Carbon & Correll, 2014; Häfner et al., 1993; Kleinhaus et al., 2011; Ochoa et al., 2012), depression (Brodaty et al., 2005; Frey et al., 2020; Szadoczky et al., 2002; Thase et al., 2005) and psychosis (Ali et al., 2022; Crumlish et al., 2009; Malla et al., 2006; Verma et al., 2012), to name a few. As such, in the following analysis, we focus on the hospitalisation outcomes of male and female patients separately.

Between-group comparisons were performed using either a two-tailed non-parametric Mann–Whitney’s U-test (for ordinal measures such as hospitalisation duration and family involvement) or a Chi-square test (for frequency measures such as primary diagnosis distribution or 30-day readmission distribution). We used SPSS version 24 (SPSS Inc., Chicago, IL) for our analysis. The p value for statistical significance was set at p ≤ .05.

Results

Our sample consists of 300 hospitalisations ranging from 2017 to 2021. Of these 300 hospitalisations, 52.3% (157) were male patients (mean age of 42 ± 13 (years ± SD)), and 47.7% (143) were female patients (mean age of 43 ± 11.5 years). These patients were cared for by 60 different residents, 23 of whom were male (mean age of 33 ± 1.9 years) and 37 female (mean age of 33 ± 1.7 years). On average, each male physician treated 4.6 ± 1.5 patients in our sample and each female physician treated 5.1 ± 1.3. The difference is not found to be statistically significant.

Table 1 presents the number of hospitalisations in the sample per physician–patient sex interaction. As can be observed in the table, in our sample, female patients were treated by female physicians more often than male patients (odds ratio (OR) of 1.4 with a 95% confidence interval (CI) from 0.9 to 2.3). However, the difference is not found to be statistically significant.

Table 1. Physician–patient sex interaction in the sample.

	Male physician	Female physician	Total
Male patient	65	92	157
Female patient	47	96	143
Total	112	188	300

As can be seen in Table 2, no statistically significant differences are encountered for either male or female patients based on the sex of their physicians. Specifically, considering the male and female patient sub-populations separately, we do not observe any significant differences in terms of patient age, age at diagnosis, number of previous hospitalisations, mean duration of previous hospitalisation in the past 5 years or mean time to readmission in the past 5 years. In addition, for both sub-populations, a roughly similar distribution of the primary diagnosis is encountered. As can be seen in Figure 1, there were no significant differences based on the physician’s sex. These results combine to support the assumption that patients in our sample are assigned to male and female residents in a quasi-random fashion.

Figure 1. Primary diagnosis per physician–patient sex interaction.

Table 2. Physician and patient mean characteristics per physician–patient sex interaction.

Physician	M	F		M	F
Patient	M	M	p Value	F	F	p Value
Age (Ph)	32.8 ± 1.6	32.9 ± 2	.81	33.7 ± 1.7	33.5 ± 1.9	.63
Age (Pa)	42.6 ± 12.9	41.3 ± 13.1	.61	44.2 ± 10.7	42.8 ± 11.5	.21
Age (Diag.)	25 ± 12.7	23.1 ± 7.8	.46	22.5 ± 5.1	23.2 ± 7	.6
#Hosp.	10 ± 6.3	11.5 ± 6.5	.11	16 ± 9.2	15.8 ± 8.2	.98
Hosp. dur.	41.3 ± 49.1	44.7 ± 45.7	.99	40.5 ± 35.6	46.7 ± 42.2	.87
Readm.	112 ± 139	181 ± 186	.25	126 ± 146	118 ± 134	.5

Each column represents one physician–patient sex interaction. Rows represent age (physician, in years), age (patient, in years), age at diagnosis (in years), number of previous hospitalisations, mean duration of previous hospitalisation in the past 5 years (in days) and mean time to readmission in the past 5 years (in days).

As one could expect, and as discussed in the introduction, male and female patients’ characteristics were found to significantly differ in our sample. For example, we find that female patients are slightly older (on average, 1.5 years older), have more prior hospitalisations (on average, five more hospitalisations), and their primary diagnosis distribution is different than that of male patients (e.g., female patients are more often diagnosed with personality and affective disorders compared to male patients, see Figure 1), with all differences being statistically significant. These results verify previous findings in the literature and further support our analytical choice of analysing male and female patients separately.

Considering the physicians’ characteristics, we do not encounter any significant differences in physician age and, as such, they are expected to have similar qualifications and experience.

We now turn to examine the hospitalisation outcomes per physician–patient sex interaction. As can be seen in Table 3, patients treated by female residents present a shorter mean hospitalisation time (for male patients: 58 vs. 66.9 days, means difference (MD) of 8.9 days, 95% CI (1, 19). For female patients: 54.1 vs. 66.7 days, MD of 12.6 days, 95% CI (4, 26)), extended time to next hospitalisation (for male patients: 269 vs. 179 days, MD of 90 days, 95% CI (26, 129); for female patients: 179 vs. 123 days, MD of 56 days, 95% CI (1, 71)), lower 30-day readmission rate (for male patients: 10% vs. 37%, odds difference (OD) of 27%, 95% CI (14%, 40%), OR of 5.4, 95% CI (2.3, 12.7), number needed to treat (NNT) to prevent one 30-day readmission is 3.7; for female patients: 19% vs. 38%, OD of 19%, 95% CI (4%, 35%), OR of 2.7, 95% CI (1.2, 5.9), NNT to prevent one 30-day readmission is 5.1), significantly higher levels of family involvement (for male patients: 2.6 vs. 2.1 points, MD of 0.5 points, 95% CI (0.1, 0.9); for female patients: 2.7 vs. 1.9 points, MD of 0.8 points, 95% CI (0.5, 1.1)), and higher chances of acquiring rehabilitation services (for male patients: 39.1% vs. 23.1%, OD of 16%, 95% CI (1.8%, 30.2%) and OR of 2.14, 95 CI (1.05, 4.37); for female patients: 34.4% vs. 17%, OD of 17.4%, 95% CI (2.6%, 31%) and OR of 2.56, 95% CI (1.07, 6.09)). All differences are significant at p < .05.

Table 3. Hospitalisation outcomes per physician–patient sex interaction.

Physician	M	F		M	F
Patient	M	M	p Value	F	F	p Value
Hosp. dur.	66.9 ± 66.1	58 ± 63.5	.03	66.7 ± 46.7	54.1 ± 43.9	.01
Readm.	179 ± 227	269 ± 194	<.01	123 ± 136	179 ± 109	.04
30d Rea	37%	10%	<.01	38%	19%	.01
Fam. inv.	2.1 ± 0.76	2.6 ± 0.72	<.01	1.9 ± 0.85	2.7 ± 0.69	<.01
Rehab.	23.10%	39.10%	.03	17%	34.40%	.03

Each column represents one physician–patient sex interaction. Rows represent hospitalisation duration (in days), time to next hospitalisation (in days), 30-day readmission rate, family involvement during stay (measured on a four-point ordinal scale: 1 = no family interaction; 2 = a single meeting with the family; 3 = several meetings with the family; 4 = complete family involvement, which includes at least one additional staff member, e.g., a social worker or a psychologist) and the acquisition of rehabilitation services.

In other words, in our data, both male and female patients, despite being statistically indistinguishable in their demographics and medical history indicators, are associated with more favourable hospitalisation outcomes (i.e., shorter mean hospitalisation time, extended time to next hospitalisation, lower 30-day readmission rate, significantly higher levels of family involvement and higher chances of acquiring rehabilitation services) when managed by female physicians.

We extend our analysis by adjusting for patient-specific hospitalisation histories. Specifically, we consider two additional measures: (1) the patient-adjusted (PA) hospitalisation duration (i.e., the difference between the current hospitalisation’s duration and the patient’s mean hospitalisation duration in the past 5 years) and (2) the PA time to readmission (i.e., the difference between the current time to readmission and the patient’s mean time to readmission in the past 5 years). These two measures are handled in two ways: first, as raw differences between past means and present values; second, as binary indicators, considering only the sign of the difference (i.e., positive or negative).

As can be seen in Table 4, both male and female patients are associated with significantly better PA outcomes (i.e., shorter PA hospitalisation duration and longer PA time to readmission) when managed by female physicians in all but one of the comparisons. In words, these patients present shorter adjusted hospitalisation periods (for male patients: 13.2 vs. 25.6 days, MD of 12.4 days, 95% CI (6, 30); for female patients: 7.4 vs. 26.3 days, MD of 18.9 days, 95% CI (10, 37)), higher chances of experiencing a shorter hospitalisation period (for male patients: 57% vs. 38%, OD of 19%, 95% CI (2.5%,33.5%), OR of 2.08, 95% CI (1.09,3.98); for female patients: 62% vs. 42% days, MD of 20% days, 95% CI (3%, 37%), OR of 2.26, 95% CI (1.1, 4.61)), extended time to next hospitalisation (for male patients: 87.2 vs. 66.5 days, MD of 20.7 days, 95% CI (–20, 73) yet the difference is not statistically significant; for female patients: 60.4 vs. −3 days, MD of 63.4 days, 95% CI (9, 87)), and higher chances of experiencing a longer time to remission (for male patients: 71% vs. 46%, OD of 25%, 95% CI (9.5%, 39.5%),OR of 2.81, 95% CI (1.45, 5.45); for female patients: 55% vs. 15% days, MD of 40%, 95% CI (26%, 54%), OR of 2.26, 95% CI (1.1, 4.61)). All differences, expect as noted, are significant at p < .05.

Table 4. Patient adjusted hospitalisation outcomes per physician–patient sex interaction.

Physician	M	F		M	F
Patient	M	M	p Value	F	F	p Value
PA dur.	25.6 ± 12	13.2 ± 19	<.01	26.3 ± 26.2	7.4 ± 26.9	<.01
Shorter dur.	38%	57%	.03	42%	62%	.02
PA readm.	66.5 ± 87.6	87.2 ± 67	.11	−3 ± 141	60.4 ± 184	.01
Longer readm.	46%	71%	<.01	15%	55%	<.01

Each column represents one physician–patient sex interaction. Rows represent the patient-adjusted hospitalisation duration (difference in days), patients who experienced a shorter hospitalisation compared to their past five year (percentage), patient-adjusted time to next hospitalisation (difference in days) and patients who experienced a longer time until next hospitalisation compared to their last five years (percentage).

Discussion

Our results seem to indicate that both male and female psychiatric patients are associated with more favourable hospitalisation outcomes when treated by female residents compared with patients cared for by male residents in the hospitalisation outcome characteristics examined in the current study. These differences include shorter mean patient hospitalisation duration, extended mean time to readmission, lower 30-day readmission rate, higher levels of family involvement and better chances of rehabilitation services the acquisition. These differences persist also when considering PA measures as well. Arguably, these differences are of clinical significance.

It is important to note that our results do not point to a clear mechanism that could explain the identified association. Prior literature has posed a number of possible explanations to account for the possible gender-based difference between male and female physicians, which could potentially apply to mental healthcare as well. These include female physicians’ tendency to follow medicine guidelines (Baumhäkel et al., 2009; Reid et al., 2010; Tsugawa et al., 2017) and engage in more patient-centred communication (Ganguli et al., 2020; Krupat et al., 2000; Roter et al., 2002). In general, women are more likely than men to engage in active listening, a technique which requires that the listener fully concentrate, understand, respond, and then remember what was said, as well as other partnership-building behaviours that may promote patient engagement (Roter & Hall, 2004). Similarly, prior research has shown that, on average, women are more emotionally intelligent than men (Schutte et al., 1998), a fact which may lead to greater expression of empathy during patient interactions and could lend to more robust patient–physician relationships. In the context of this study, one could argue that higher levels of family involvement and higher chances of acquiring rehabilitation services, both better associated with female residents, are potentially explanatory for other hospitalisation success measures (e.g., lower 30-day readmission rate). However, further study is required in order to determine how these and other possible factors such as clinical acumen and teamwork could explain the recorded differences. We plan to carry out such an inquiry in the future.

This study has several limitations that offer additional fruitful avenues of future research. First, our sample is taken from a single mental healthcare centre. Official data from the Israeli Ministry of Health (Israeli Ministry of Health, 2019) show that Shalvata is roughly representative of the Israeli mental healthcare system in almost every examined measure (e.g., patient characteristics, hospitalisation outcome, etc.) for at least the last decade. As such, we plan to extend our investigation outside of Israel in the future. Second, as is the case in prior studies in this line of inquiry (e.g. Greenwood et al., 2018; Tsugawa et al., 2017; Wallis et al., 2017), it is possible that omitted or confounding variables (which are not measured or accounted for in this study) may correlate with physician–patient gender concordance and patient outcomes. For example, it may be the case the female physicians work closer with female nurses and other staff members (e.g., social workers) who, in turn, bring about better outcomes than their male counterparts. Conversely, it may be the case that other work related practices and perceptions, in which female residents differ, are the contributing, in part, to the identified differences. We plan to examine these and similar factors in future work. Third, lastly, an important factor to consider in future work is the effects of supervising physicians and departmental officials on physicians success and patient outcomes.

In conclusion, our study suggests that, in accordance with previous studies in general medicine, hospitalised patients treated by female psychiatry residents present significantly better hospitalisation outcomes compared to those treated by male residents. Further work is needed in order to investigate the sources and reasons for the identified differences.

Ethical approval

This study was approved by the Shalvata Mental Healthcare Centre’s Institutional Helsinki Board, approval SHA-0003-22.

Consent form

Patient informed consent was waived by the Shalvata Mental Healthcare Centre’s Institutional Helsinki Board.

Disclosure statement

The authors declare that they have no conflict of interest.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.

Notes

1 We use the term "quasi-randomly" since a true randomised allocation mechanism cannot be implemented in practice given residents’ exams, vacations, sick-days, etc.

2 All patients admitted to one of the closed wards during 2021 were extracted and randomly sorted. The first 300 patients who met our criteria were selected.