Psychiatric Hospitalizations: A Comparison by Gender, Sociodemographics, Clinical Profile, and Postdischarge Outcomes

We performed a population-based cohort study of all individuals discharged from inpatient general psychiatric units in Ontario, Canada (2008–2013). Ontario is Canada’s largest province (population of approximately 14 million). Its demographic characteristics are largely nationally representative, with about 90% of residents living in urban areas and about one-quarter being non-Canadian born. The health care system is government funded, and it covers all Ontario residents and includes mental health care. General psychiatric units provide most inpatient psychiatric services for individuals with mental illness in Ontario. Most patients are admitted after an emergency department assessment, but some patients are directly admitted via their treating physician. We included only unscheduled hospital admissions in the study reported here.

We used linked population health administrative databases at Ontario’s Institute for Clinical Evaluative Sciences (ICES), an independent, nonprofit research organization that holds population-level data to evaluate health care services and their effectiveness. Patient-level records in these data are linked to each other with a unique encoded identifier for every Ontario resident who has an assigned health card number. The Ontario Mental Health Reporting System (OMHRS) database comprises all hospital admissions for adults ages ≥18 in psychiatric inpatient beds and includes the Resident Assessment Instrument–Mental Health (RAI-MH), a comprehensive clinical assessment tool. The RAI-MH includes patient sociodemographic characteristics, DSM-IV diagnoses, and measures of psychiatric symptoms, substance use, cognition, and functional impairment. All items have adequate reliability (interrater agreement >80%) and validity in inpatient settings (8). Data on medical hospitalizations were obtained from the Canadian Institute of Health Information discharge abstract database. We used the Ontario portion of Canada’s National Ambulatory Care Reporting System for data on emergency department visits and the Ontario Health Insurance Plan for physician billing claims data for outpatient services.

We considered all psychiatric inpatients older than age 18 in Ontario (from April 1, 2008, to September 30, 2013). We included only patients who were hospitalized for >72 hours, as in our previous work, to focus on those hospitalized long enough to receive inpatient treatment and to have had an opportunity to arrange postdischarge care (2). We considered the first admission per individual in the study period as the index admission. Sex (male or female) was identified on the basis of patient self-report in the OMHRS data set. We also extracted sociodemographic characteristics, including age, marital status, living situation and type of residence, source of income, educational attainment, and life stressors (including history of abuse or assault). We linked participant postal code to Canadian Census data to derive average neighborhood income for each individual, dividing this into distribution quintiles to obtain estimates of relative neighborhood income. We also used postal codes to distinguish rural (population <10,000) from urban areas. Mental health services and diagnoses before index admission were obtained from inpatient and outpatient data sources. Validated ICES disease registry data sets were used to identify diabetes, hypertension, and chronic obstructive pulmonary disease, and overall general medical comorbidity was estimated with a Charlson Comorbidity Index score (9).

Characteristics of the index admission included involuntary admission, concerns about harm to self and harm to others, length of stay, planned versus unplanned discharge (that is, leaving against medical advice), DSM-IV psychiatric diagnoses, and Global Assessment of Functioning (GAF) score (8,10). We considered each of five RAI-MH symptom scales—depression rating scale, positive symptom scale, negative symptom scale, mania symptom scale, and cognitive performance scale—and summed six data elements representing anxiety symptoms (see the OMHRS coding manual for details [11]). We also considered RAI-MH scales that assess function: activities of daily living hierarchy scale, aggressive behavior scale, risk of harm to others scale, severity of self-harm scale, and the self-care index (8). Admission scales were administered within three days of admission, and discharge scales were administered at discharge.

Postdischarge psychiatric outcomes, measured within 30 days and 90 days after discharge from the index admission, were readmission, emergency department (ED) visits without hospital readmission, self-harm (with or without a readmission), and death. Readmission was either any readmission to a designated mental health bed or any readmission to an acute care hospital nondesignated hospital bed where the primary diagnosis was a mental health condition (ICD-10-CA codes F00–F99), including admissions to intensive care units after suicide attempts and to medical beds when designated psychiatric beds were unavailable. Emergency department visits without readmission included those in which the primary diagnosis was for a mental health condition (ICD-10-CA codes F00–F99) or in which deliberate self-harm was recorded in any diagnostic field (ICD-10-CA codes X60–X84, Y10–Y19, and Y28) (12). Self-harm was defined as an emergency department visit (with or without readmission) with an emergency department code for deliberate self-harm as noted above (12). All-cause death was captured with the enriched Ontario Registered Persons Database.

We used standardized differences to compare cohort characteristics between men and women. Standardized differences compare differences in units of the pooled standard deviation and are not influenced by sample size. Standardized differences of .1 (that is, a 10% difference) denote a meaningful imbalance on the variable (13). Odds ratios (ORs) and 95% confidence intervals were calculated to compare women and men on outcomes at 30 and 90 days after discharge, with men as the reference group. Models were adjusted for baseline characteristics that differed between women and men. Subanalyses were conducted in which the main analysis was repeated among cohort members with major depressive, bipolar, psychotic, and alcohol or drug use disorders. SAS version 9.3 for Unix was used for analysis. Cell sizes <6 are not reported for privacy reasons. Research ethics board approval was from Sunnybrook Health Sciences Research Centre (ICES study registered as 20120584425000).

Of 95,055 admissions, 48,083 patients were women (50.6%) and 46,972 were men (49.4%). Women were older and had greater socioeconomic stability but were more likely to report life stressors, including recent and remote trauma. Men and women had similar recent and lifetime rates of psychiatric admission, emergency department use, and self-harm attempts, and general medical comorbidity was similar for men and women. Men were more likely than women to present with psychotic disorders, alcohol use, and substance use disorders and to be assessed as aggressive toward others. [A table in the online supplement to this report summarizes data on these and other characteristics.] Women were more likely to have mood and anxiety disorders and comorbid personality disorders. [See table in the online supplement.] Length of stay was similar between men and women, and a minority (<10%) in both groups were discharged against medical advice. Both men and women showed symptom improvement during the admission, but women had higher depression and anxiety symptoms at discharge. Characteristics of admitted patients were consistent throughout the study period.

Postdischarge outcomes are shown in Table 1. About 9.0% of the cohort was readmitted by 30 days postdischarge, increasing to 15.8% by 90 days after discharge, with no differences between men and women after accounting for clinical factors. Overall 6.3% and 11.1% of the cohort had an emergency department visit without readmission within 30 and 90 days of discharge, respectively, with similar risk for men and women after covariate adjustment. Within 30 days of discharge, 1.3% of the cohort had a self-harm attempt, and by 90 days postdischarge, this increased to 2.5%. Women were significantly more likely than men to harm themselves (AOR=1.17 for 30 days and AOR=1.28 for 90 days). Death was rare (<1% at both time points for men and women), but women were at lower risk (30 days, AOR=.49; 90 days, AOR=.53). [Subgroup analyses by primary diagnoses are shown in a figure in the online supplement to this report.]

In a population-based study of psychiatric inpatients, we found important differences suggesting that men and women may require different types of intervention to address their social determinants of health and clinical needs during hospital stays and postdischarge.

Although the relative socioeconomic stability of women could be considered advantageous, their rates of revisits to the emergency department and of readmission were similar to those of men, and women’s self-harm rates postdischarge were higher than men’s. Most existing discharge planning and postdischarge transitional care initiatives focus on more basic socioeconomic needs, such as maintaining housing stability, offering case management to ensure attendance at appointments, and pharmacotherapeutic adherence (14). However, our results suggest that women were more likely than men to have these basic needs met, and so, to improve postdischarge outcomes, women may require other services—for example, more social and practical support—to address their recovery in the context of their roles as wives and mothers.

This could be especially important given the elevated rates of recent physical and sexual abuse in the sample, as well as the higher proportion of women reporting child custody issues as a recent life stressor [see online supplement]. The lifetime history of trauma was 1.5 to 3 times higher among women compared with men in our cohort, which is consistent with existing literature (15) and may have implications for how services might be tailored on inpatient units and how groups of male and female patients should be integrated on these units. To our knowledge, no randomized controlled trials have explicitly evaluated the effects of trauma-informed inpatient care (or gender-specific programming) on inpatient or postdischarge outcomes such as readmission (14). Of note, many male patients also reported a history of trauma, and so may also benefit from consideration as to how trauma affects their experiences and prognoses related to psychiatric inpatient hospitalization.

On the whole, outcome differences in terms of the higher risk of self-harm and death among women are consistent with prior literature, as are the lack of differences in rate of return to the emergency department and hospital (2,4,7). However, the high rates of emergency revisits and readmissions in the 30 and 90 days postdischarge support continued vigilance for both men and women during this period, especially since these outcomes are costly to the system and disruptive to the patient recovery process.

A major strength of this study was its use of comprehensive clinical and health administrative population-level data. We did not investigate mechanisms underlying the adverse outcomes postdischarge, such as nonadherence to medication, problems related to housing, or substance use relapse, however. Nor did we generate sex-specific models of independent predictors of postdischarge outcomes—for example, evaluating the impact of social factors such as trauma history or clinical factors such as symptomatology and inpatient treatment regimens. We did not analyze cause of death in our sample, but given the relatively young, healthy cohort, the excess deaths among men were likely due to external causes of injury (such as suicides or accidents) instead of medical conditions. Of note, we limited our study to index admissions lasting ≥72 hours, so our findings may not be applicable to shorter-term crisis-type admissions. As with most large population-based studies, clinicians should be mindful that statistically significant differences do not necessarily imply that differences are clinically significant.

Men and women hospitalized on psychiatric inpatient units had distinct sociodemographic characteristics, distinct clinical presentations, and some differences in postdischarge outcome patterns. Future research is needed to determine whether these data can inform development of targeted supports and services to help deliver more person-centered care to improve outcomes for both men and women.