Continuity of Care and Health Care Costs Among Persons With Severe Mental Illness

Continuity of care has long been considered an important characteristic of services for persons with severe mental illness, yet empirical evidence for an association between continuity and health outcomes is limited (1). In addition, despite claims that continuity of care is efficient (2), little is known about its relationship with health care costs. This is an important gap in the literature, because policy makers must decide whether allocation of resources to services that focus on enhancing continuity make sense in terms of health outcomes as well as costs to achieve those outcomes.

Saarento and colleagues (3) have called for further studies in this area and note that such studies "should include outcome measures, cost analyses, and information about the local healthcare system and its functioning as a whole." It has been argued that if continuity improves patients' well-being, then hospitalizations and use of other health care services are likely to decrease (4). Decision makers should thus be interested in determining first whether an association exists between continuity and health care outcomes, and, subsequently, whether an association exists between continuity and costs. This latter relationship has been examined in various fields, including among family practice patients, for whom provider continuity has been found to be a significant predictor of health care costs (5), and in the case of chronic nonpsychiatric conditions, for which reductions in use and costs with improved continuity have been reported (6). However, limited data are available for the mental health field.

For example, Chien and colleagues (4) found that improved continuity was associated with lower Medicaid costs, although their study was limited by a narrow definition of continuity and examined only patients with schizophrenia. In a study of veterans in the United States, researchers found that outpatient expenditure was a predictor of mental health care continuity (7), but, again, a relatively narrow construct for continuity was used. In addition, specific interventions that are intended to improve continuity, such as assertive community treatment teams, have been shown to reduce costs (8,9,10). The focus of the study reported here was to examine a more comprehensive measure of continuity at the system level and to identify relevant relationships with health outcomes and cost.

Given Canada's publicly funded, universal access, single-payer health care system, we had an opportunity to study continuity of care and its association with health care outcomes and costs across a full continuum of services in a longitudinal cohort of patients with severe and persistent mental illness. Here we report on the relationship between continuity of care and health care costs in a sample of 437 patients in three health regions in Alberta, Canada. The specific aims of the study were to report on observed associations first between continuity and cost and, second, between cost and outcomes. A companion paper in this issue of Psychiatric Services (11) presents findings on continuity and health care outcomes. We hypothesized a priori that continuity of care would be associated with improved patient outcomes and reduced health care costs.

Methods

The cohort study

The study used an observational cohort of 486 patients who were followed over a 17-month period (March 2001 to December 2002) in three health regions in Alberta, Canada. Complete cost data were available for 437 patients. The design and methods of the cohort study are described in more detail in the companion article in this issue (11). Eligibility for inclusion of patients with severe mental illness was confirmed by using a structured diagnostic interview, the Mini International Neuropsychiatric Interview (MINI), and a clinician-confirmed two-year history of the condition. The study was reviewed and approved by the conjoint health research ethics board at the University of Calgary. Informed consent for prospective data collection was gained at study entry, and a separate consent for access to government administrative data was affirmed in writing by each participant before the data were released.

Continuity and outcome measures

The focus of this analysis was on associations between continuity of care and various service costs. To keep the analysis manageable, we used the observer-rated component of the Alberta Continuity of Services Scale for Mental Health (ACSS-MH) as our independent variable for these analyses. This instrument is described in more detail in the companion article (11). As a secondary aim, we also examined the associations between costs and outcomes, with continuity retained as a predictor variable.

We focused this analysis on the outcome measure that had shown the strongest associations with continuity in the main study (the EQ-5D) (12). We also added an observer-rated outcome measure of community functioning, the Multnomah Community Ability Scale (MCAS) (13), to explore whether patterns of association found with self-reported outcomes would persist with observer-rated outcomes.

Utilization and cost measures

The study was conducted from a payer perspective—that is, government health care expenditures. Administrative data that included hospital-based inpatient, outpatient, and emergency service events were received directly from the records department in each of the three health regions. In the smaller of the three regions, patient-specific costs were supplemented with per diem rates because of a lack of availability of the former. In total, less than 5 percent of all inpatient stays were costed on a per diem basis. Number of days of care in specialist psychiatric facilities (one in each region) was captured through administrative data sets from the Alberta Mental Health Board. Per diem rates for the admitting unit were used because patient-specific costing was not available for these facilities. In total, this type of costing applied to 25 visits (less than 1 percent of all hospitalizations) across the entire sample. Physician billing data, including community and hospital-based visits by family physicians and specialists, were obtained from Alberta Health and Wellness on a patient-by-patient basis. Only hospital and physician events that were determined to be attributable to a mental illness by clinical co-investigators (psychiatrists) through a priori defined decision rules (available on request) were included in the cost analysis.

Use of community-based services—including clinic and crisis care, indirect treatment agencies, and residential, vocational, addictions, and housing services—were tracked by using patient interviews at two- to three-month intervals. Standard provincial rates based on the provincial collective bargaining agreement were used for all nonphysician professionals (14), and relevant agencies were surveyed for cost estimates for their services. Social workers, psychologists, mental health therapists, vocational counselors, registered nurses, and registered psychiatric nurses were included, whereas dentists, opticians or optometrists, audiologists, naturopaths, homeopaths, chiropractors, herbalists, acupuncturists, massage therapists, and physiotherapists were excluded. Benefits of 19 percent were added to all hourly wages.

Home care and outreach services were recorded and costed at a flat rate according to regional service data, with $.33 per kilometer added on top of hourly rates. Crisis mobile team visits were included and were costed on the basis of local costing information. A rate of $1 per call was attributed to volunteer-staffed call centers—for example, a distress line. Patients receiving government income support (Assured Income for the Severely Handicapped) of $855 per month were identified, and out-of-pocket drug costs were recorded, but neither were included in the analyses reported here. Almost all the patients in this sample had government assistance, which would cover at a minimum 80 percent of community-based drug costs. All drugs administered in the hospital are covered by the government as part of the inpatient stay.

Standard service and medication summaries were produced on the basis of patient information provided at follow-up and on chart review and were costed on the basis of provincial rates (15). Dispensing fees were added on the basis of Alberta guidelines (16). When no information on dosage was provided or when such information appeared to be inaccurate in light of the a priori defined decision rule (extremely low or high values as per the Compendium of Pharmaceuticals and Specialties [CPS]), the average recommended dose (as per the CPS) was applied. Such adjustments occurred in less than 2 percent of all cases. All drug costs reported include outpatient and community drugs; inpatient costs include inpatient-administered drugs. Community laboratory tests were also included and were costed on the basis of provincial cost lists (17). Only drugs and tests that were deemed by the clinical coinvestigators to have been prescribed or ordered for a mental illness were included, the exception being one general blood screen per year, which was included as standard care for each patient. The cost elements are summarized in Table 1. All costs are reported in 2002 Canadian dollars ($1CAN=$.81U.S.).

Analysis

First, mean±SD patient costs, along with median costs, for the 17-month follow-up period (as well as annualized values) are reported for the overall cohort on the basis of five cost categories: costs resulting from hospitalization, costs resulting from use of community services (all community services including general practitioner physician costs), drug costs, non-general practitioner physician costs (outpatient and inpatient, including psychiatrist costs), and total costs (sum of these four cost variables). Total payer costs for the entire sample over the study period are also presented.

Second, one-way analysis of variance was used to examine differences in means of log- and natural log-transformed data for the five cost categories across quartiles of observer-rated continuity of care. Although our measure of continuity is a continuous variable, reporting in quartiles enabled associations to be described in terms of a set of readily defined and practically meaningful levels. Multiple linear regression was then used to examine associations between continuity of care (the independent variable) and total, hospital, and community cost (the dependent variables), with adjustment for potential confounders (age, household income, duration of illness, recruitment location, and suicidality) (not reported). Only variables that met the standard epidemiologic definition for confounding were included in the regression analysis (18).

Third, one-way analysis of variance was used to examine the difference in means of transformed cost data across quartiles of two outcome measures, the EQ-5D visual analogue scale (a self-rated quality-of-life scale) and the MCAS. Ordinal regression analysis was then used to examine associations between MCAS and EQ-5D scores (dependent variables) and total, hospital, and community costs (independent variables), controlling for potential confounders (household income, previous hospitalization, and suicidality for the EQ-5D visual analogue scale and recruitment location within region, household income, previous hospitalizations, and illness severity for the MCAS). The significance level for all statistical testing was .05.

Results

A total of 486 patients were enrolled in the cohort study; cost data were available for 437 (90 percent) of these. The remaining patients either withdrew from the study, were lost to follow-up, or had incomplete cost data. Mean total cost, hospital cost, community cost (including general practitioner costs), drug cost, and non-general practitioner physician cost (including psychiatrist costs) were $24,070±$25,643, $12,505± $20,991, $2,848±$4,420, $4,867± $6,825, and $4,030±$5,081, respectively. Keeping in mind that the cost data were highly skewed, the median values, in the same order, were $15,014, $3,935, $1,447, $2,885, and $2,299. The total cost to the system for this cohort of 437 patients over the 17-month study period was just over $10.5 million (or approximately $7.4 million per annum).

Mean costs across quartiles of observer continuity of care are reported in Table 2. The difference in means across levels of continuity were not statistically significant for total cost but were significant for the other four cost categories. There was a $3,014 difference in average community costs from the lowest to the highest level of continuity and a $4,303 difference for average hospital costs. Median differences across levels of continuity were $1,511 for community costs and $4,790 for hospital costs. The relationship between the total, hospital, and community cost variables and continuity was examined in separate regression models, with potential confounders controlled for. As is shown in Tables 3 and 4, the indicated bivariate relationships between continuity and both hospital cost and community cost remained.

The mean cost across quartiles of the EQ-5D visual analogue scale and the MCAS are presented in Table 5. Total cost and non-general practitioner physician cost were significantly lower for patients who had a higher self-rated quality of life as indicated by the EQ-5D visual analogue scale. In addition, patients with higher rated MCAS functioning scores had significantly lower total, community, and non-general practitioner physician costs. Regression analysis revealed that none of the cost variables were significant predictors of EQ-5D visual analogue scale scores once potential confounders were controlled for (models not shown). However, both total and community cost variables were significant predictors of MCAS scores (models not shown).

As reported in the companion article in this issue (11), associations between scores on the observer-rated scale and generic quality of life (including ratings across five areas of quality of life and a total rating out of 100) and the patient scale and community functioning were statistically significant at the .001 level. These patterns suggest that there is a relationship between strong continuity of care and positive health outcomes.

Discussion

The results of this study highlight two key findings. First, poorer continuity of care is related to higher hospital costs and lower community costs, or, conversely, better continuity is related to lower hospital costs and higher community costs. Thus there is an indication that greater community investment is associated with improved continuity and that the patients with better continuity may in turn be using fewer hospital services. Second, as would be expected, patients with better functioning appear to cost the system less in total and, in particular, have lower community costs.

However, it must be stated that the direction of causality with these findings is unknown. With the observational design used in this study, it is not clear whether the reported cost findings are due to the level of continuity or, conversely, expenditure is in fact driving continuity. Because this is the first study to examine the relationship between continuity and costs at the system level, it will be necessary to conduct further research, particularly research using experimental designs, to elucidate the direction of the reported relationships. What is known from the results reported in this article is that a series of associations do exist.

In comparison, in a recent study in the United States, greater continuity of care was associated with lower Medicaid costs among patients who had two or more mental health visits in one year and was also associated with lower likelihood of psychiatric hospitalization (4). However, the same study showed that continuity was not associated with general life satisfaction or satisfaction with health. Adjustments were made for a large range of factors, leading to a potential concern about overadjustment. In addition, previous work has shown that per capita outpatient expenditure is a predictor of continuity of care (7). Other studies have reported both cost of illness and predictors of cost in mental illness (19,20,21,22), and some have examined forms of service organization in relation to cost for this population (23). However, to our knowledge none have attempted to describe costs in relation to continuity and outcomes for patients with severe mental illness across a full continuum of services. As such, these results, although descriptive, represent an important step.

As shown in the companion article, and reiterated in the Results section above, better continuity is associated with better patient outcomes. In addition, two associations were delineated here: that better continuity is related to lower hospital costs (but that there is no difference in total cost) and that higher-need patients (those with lower functioning) cost the system more. Studies with experimental designs are needed to examine whether improving continuity for persons with high needs and high costs produces better outcomes if system costs can be averted away from inpatient hospitalizations and emergency visits. The hypothesis is that improving continuity would be good for the patient and could also be seen as a mechanism for shifting costs from hospitals to the community.

If this hypothesis were to hold true, through further research, clinicians and policy makers would then need to consider developing strategies to target high-need patients with severe mental illness for interventions that improve continuity of care. Of course, achieving continuity of care comes at an economic cost—namely, a requirement for greater community expenditure. However, given that total costs in our sample did not change across levels of continuity, the cost shifting may, in effect, produce better outcomes with no increase in expenditure overall. In economic terms, this is known as the technically efficient service delivery option, and, all other things being equal, it should be pursued (24).

This study did not address important issues pertaining to allocative efficiency—that is, whether resources should be invested in patients with severe mental illness in the first place, as opposed to other potential uses of limited health care resources (24). Our study highlights the need to further examine the notion of shifting resources from hospital services to community services for this group of patients. Further work is also required to examine outcomes obtained through improvements in continuity relative to the benefits obtained through treatment of other patient groups.

Although this study is the first of its kind to assess costs across the full continuum of care in relation to both continuity and outcomes for patients with severe mental illness, several limitations should be highlighted. First, as stated above, because this was an observational study, caution must be taken in concluding causality. Experimental designs are required to examine the direction of the reported relationships.

Second, there may be a problem of endogeneity in that service visits are one aspect of the measure of continuity and also affect patient-level costs. Although this is an inherent and vexing problem—as any measure of continuity will likely include number of visits as part of its construct—it is also the case that the 17 objective items of the ACSS-MH go far beyond simple visit-based measures (25). In fact, our measure of continuity can be viewed as a substantial improvement on the continuity measure used in the only other previous empirical examination of the relationship between continuity and cost (4,7).

Third, this study adopted a payer perspective. Ideally, both government- and patient-borne costs would be included in economic evaluations (26). In our case, out-of-pocket costs were collected, but these data—in particular, data related to lost productivity—were found to be unreliable because of collection discrepancies across the three health regions.

Fourth, the level of detail in the cost data available at the centers in this study probably complicated our data to some extent. This may be masking certain associations and plausibly explains the relatively poor fit of our models. That said, the Canadian setting afforded us the opportunity to compile cost data from every service used across the full continuum of care.

Fifth, although we relied on self-report in some instances in order to obtain activity and costs on a complete range of services, recent work has shown that patients with severe and persistent mental illness can reliably provide such information (27). In addition, we used methods to prompt patient recall and were able to triangulate reporting through the use of administrative databases.

Conclusions

This study applied a new comprehensive measure of continuity of care for persons with severe and persistent mental illness in a cohort of patients across three health regions in Alberta. The companion article in this issue of Psychiatric Services (11) reports key associations between continuity of care and health outcomes. In this paper we have presented a number of relationships between continuity of care and cost, and cost and outcomes. Data limitations notwithstanding, this work has provided an important step in quantifying costs for this patient group.

We have shown a promising relationship between levels of continuity and both hospital and community costs. The data further indicate that a relationship exists between cost and level of patient functioning. Taken in conjunction with the continuity and outcomes results, these findings point toward a need for more work in this area to examine whether shifting resources from hospital to community, particularly for high-need patients, would result in improvements in continuity of care and subsequent health outcomes.

Acknowledgments

This study was funded by grant RC2-2709 from the Canadian Health Services Research Foundation and cosponsored by the Alberta Heritage Foundation for Medical Research, the Alberta Mental Health Board, the Institute of Health Economics, and Eli Lilly (Canada) Inc. (unrestricted grant). This study is based in part on data provided by Alberta Health and Wellness. The authors thank Doreen Ma, M.Sc., and Lynne Kostiuk, M.Ed. The interpretation and conclusions contained herein are those of the researchers and do not necessarily represent the views of the Government of Alberta or Alberta Health and Wellness.

Dr. Mitton is affiliated with the department of health care and epidemiology of the University of British Columbia in Vancouver. Dr. Adair is with the department of psychiatry and the department of community health sciences of the University of Calgary. Dr. McDougall is with the department of psychiatry of the University of Calgary and the Alberta Mental Health Board. Ms. Marcoux is with the department of community health sciences at the University of Calgary. Send correspondence to Dr. Mitton at the Centre for Healthcare Innovation and Improvement, 4480 Oak Street, E414A, Vancouver, British Columbia, Canada V6H 3V4 (e-mail, [email protected]).

Enlarge table

Enlarge table

Enlarge table

Enlarge table

Enlarge table