Variation amongst hospital medicine programs in Canada: development and validation of a classification scheme and diagnostic tool

Vandad Yousefi, Mark Evans

Abstract

There is significant variability across Canada in how hospitalist programs are structured to deliver inpatient care. As a result, it is very challenging for hospital managers and physician leaders to evaluate hospitalist program structures, processes and outcomes against meaningful benchmarks. We present our research into the development and validation of a classification scheme, and show that programs can be categorized into clusters based on a number of key characteristics. This allows hospital managers to identify peer programs and establish a set of best practices based on where their hospitalist program sits on a continuum of increasing complexity.

Introduction

Hospital Medicine (HM) in Canada has experienced a consistent growth since the late 1990s. Hospitalist programs are now present in community hospitals of various sizes and increasingly in large academic centres. Although many of the drivers for the creation of hospital medicine programs are similar across the country (such as the exodus of family physicians from delivering hospital-based Most Responsible Provider care) (1),  the majority of programs have grown independently and in response to local needs for inpatient care coverage. As a result, the design and operations of these programs have been heavily influenced by the local contexts within which they have evolved. For example, while hospitalists in some programs are responsible for a large array of patients with high acuity levels, in some jurisdictions their scope of practice is limited to a less acute patient population. This variation in program design and operations also extends to other key areas of the hospitalist model such as governance (eg. a distinct department of Hospital medicine vs. a program/division within other departments), coverage model (24/7/365 coverage vs. daytime only), workload (high vs. low individual patient census) and compensation (fee-for-service vs. alternative payment models).

We have previously described a typology to classify HM programs into several distinct categories along a continuum of increasing program complexity (Table 1) (2). This work was based on our extensive qualitative analysis of various HM programs and our consulting work in helping organizations develop new programs or optimize existing ones. We described three stages in the evolution of hospital medicine in Canada: the traditional model of inpatient care in community hospitals (the “pre-hospitalist” stage) where primary-care providers looked after their own patients; the “partial hospitalist” stage in which community-based providers formed groups looking after a pool of patients (such as “doctor of the day” programs); and the “hospitalist” stage where a defined group of providers primarily dedicated their time and clinical focus to hospitalized patients. Within this stage, we also described several “generations” of hospital medicine programs, with varying degrees of structural complexity and increasing footprint in their host institutions.

In our experience, many hospital medicine programs initially start out as rudimentary groups of primarily community-based providers who come together to cover a growing number of unattached patients. As demand grows, hospitals implement more formal inpatient care programs, usually drawing on the existing pool of physicians engaged in inpatient care at the facility. With time, these programs evolve into formal divisions or departments, with increasing numbers of individuals with a dedicated focus on hospital-based care looking after an increasing number of hospitalized patients. As programs evolve, many may start providing supportive care to other providers (surgeons, psychiatrists) and engage more in non-clinical activities such as teaching, research and quality improvement.

While this typology makes intuitive sense to us and has been validated over the years through direct experience and observations of a large array of hospitalist programs across Canada, we have not come across any formal attempts, to date, to study the issue of hospitalist program variation in the Canadian context. In order to validate our typology, we used the results of the most recent national survey of Canadian hospitalists (conducted by the Canadian Society of Hospital Medicine - now the Canadian Chapter of the Society of Hospital Medicine - in 2012) to determine if distinct types of hospitalist programs can be identified based on objective characteristics. In this article, we outline our analysis of the survey results and validation of our previous classification scheme, and outline some policy implications for researchers and health system leaders who are interested in how inpatient care programs can be optimized to deliver desired outcomes. 

Methods

Survey design and target population

We collaborated with the Canadian Society of Hospital Medicine to design and implement the 2012 National Hospital Medicine Survey (NHMS). The survey was based partly on prior similar surveys conducted in Canada and the United States of individuals who self-identify as hospitalists (3, 4). It was designed to provide insights on various aspects of individual hospitalists’ work such as workload, scheduling model, compensation, and professional satisfaction. The survey also contained detailed questions on characteristics of the broader hospital medicine programs to which the individuals belonged, such as scope of practice and services provided (eg. surgical co-management, resuscitation services), program design (eg. dedicated program leadership, governance structure, types of providers involved), program size (eg. number of full time equivalent hospitalists, number of inpatient beds covered by the program, daily census), and some outcome metrics (eg. self reported length of stay, involvement in quality improvement projects, physician satisfaction). The survey targeted self-identified hospitalists who had agreed to share their contact information with the CSHM. It was conducted online in July 2012, with 1320 individuals invited to participate. A total of 305 responses were received (response rate 23 %), representing 80 hospitalist programs in Canada.

Statistical Analysis

In order to identify potential differences between HM programs, responses from individuals working in the same institutions were aggregated to program level. Latent class segment modelling were applied to identify unique groups that were sufficiently different based on statistically significant differences in relevant measures. We considered various institutional characteristics (such as scope of practice, program size, number or providers etc.) and used parametric and non-parametric regressions to identify statistically significant differences. We subsequently calculated the probability that each of the 80 programs represented in the NHMS belonged to these segments. 

Results

 Our latent class modelling identified four segments amongst the HM programs in our sample population that generally correspond to our previous classification scheme. Given that the survey population comprised of individuals who self-identified as being involved in providing hospitalist care, traditional inpatient care programs (“pre-hospitalist” stage programs) were not represented in our sampling and as a result did not appear in our segmentation. The four identified segments correspond to “partial hospitalist” and the three generations of “hospital medicine” programs. Table 2 outlines key differences between the segments in various program characteristics that broadly confirm our initial descriptions. For example, programs clustered in Segment D (Corresponding to Third Generation HM programs) have the highest number of full time providers, higher patient volumes, a more formal contractual relationship with their hospitals, and more committee involvement. These programs are primarily found in larger institutions, and providers expressed higher satisfaction with various aspects of their practice. On the other hand, unlike our previous predictions, Segment D programs have a more focused scope of practice with lower levels of participation in procedures or critical care coverage. This is likely explained by the fact that in larger institutions, other subspecialty services (such as dedicated intensive care physicians) exist which obviate the need for hospitalist involvement in the care of specific segments of hospitalized patients.

 We found five variables that discriminated between the four clusters identified by our analysis: type of provider group and nature of group organization, the type of relationship between the physician provider group and the institution, scope of clinical services, presence of a formal team/program leader, and percentage of hospital beds covered by the HM program (Table 3).

 The four segments identified in our study population can be arranged along a continuum that represents progressive changes in various aspects of program structural and operational characteristics (Figure 1). For example, there is a clear trend with respect to the length of time programs in various clusters have been in operation, confirming our earlier hypothesis that hospitalist programs can mature and change over time to more complex care delivery systems. The modelled outcomes and program benchmarks also provide a directional framework with which program managers can begin to understand the implications and relative merits associated with program type and organizational transitions from one segment to another.

 Discussion

 Although hospitalist programs have existed in Canada for over 20 years, as far as we can determine our study is the first formal attempt to make sense of the observed variability in programs by identifying “clusters” with similar characteristics. Previous research has focused on classifying MRP groups based on provider/physician characteristics (4). For example White and colleagues found that individuals who provided in-patient care in Ontario could be grouped into 3 main categories based on the volume of services they provided: comprehensive community practitioners who primarily focused on outpatient work, “mixed-practice” physicians who engaged in both outpatient and inpatient work, and primarily hospital-based physicians (“hospitalists”) who focused primarily on inpatient care (on either a full time or part time basis). Our study is different as it focuses on program level attributes, taking into account both “structure” and “process” characteristics.

Our findings confirm our previous typology for classifying hospital medicine programs in Canada along a continuum of increasing complexity and evolving service delivery attributes. They suggest that despite significant observed variability amongst hospitalist programs in the country, distinct clusters can be identified based on a number of key program characteristics. While each program has evolved in response to unique local institutional needs, programs grouped in each cluster have sufficient similarities that can distinguish them from those in other segments, allowing for identification of peers and ultimately facilitating benchmarking. This is particularly useful for hospital administers who plan to implement new programs, or those who are looking to find ways to determine benchmarks for expected outcomes based on peer programs.

The ability to classify HM programs can also potentially allow for development of clinical standards in areas such as scope of services, provider competencies, and program resource requirements, as well as a framework for application of existing national standards. For example, the Society of Hospital Medicine has outlined an inventory of 47 key characteristics for an effective HM program (5). However the ability to apply these characteristics to a given HM program may depend on where the program lies on its growth trajectory. Understanding which “cluster/segment” a program belongs to can allow administrators to identify relevant best practices. For example, program administrators may identify the need for a dedicated HM program lead as an important requirement amongst programs belonging to a specific cluster. Similarly, a clear Most Responsible Provider policy that clarifies admission criteria is more common amongst programs in the Third Generation cluster, suggesting that the presence of such policies can be important for programs in larger institutions.

Our findings could have a number of policy and healthcare planning implications. They confirm that given the wide variability observed between HM programs, a “one size fits all” approach to issues of compensation and scope of services is unlikely to succeed in effectively managing hospitalist issues. This may be particularly relevant in jurisdictions in Canada where hospitalist compensation and program implementation are determined on a provincial level (such as British Columbia and Alberta). An approach that aims to utilize our proposed typology and assign a given HM program into a smaller segment is likely to provide a better opportunity for identifying peer groups and addressing common issues that affect a smaller cohort of HM programs. For example, third generation programs may require a different compensation arrangement compared to those that belong to other categories.

Our findings also highlight the need for a better understanding of the impact of physician organization on inpatient care outcomes. There is a growing recognition for the impact of inter-professional team-based care on outcomes for hospitalized patients (6). Similarly, there are many examples of quality improvement interventions that focus on optimization of care processes (such as the use of lean methods, implementation of care transition bundles etc). However, little attention has been paid to how these efforts are impacted by the way physician providers are organized in groups as part of integrated inpatient care programs, and how health system leaders and managers can facilitate better integration of hospital-based physicians in broader inpatient care teams. Future research should focus on discovering if there are any differences in outcomes among the various types of hospitalist programs and a more in-depth understanding on how various structure and process attributes relate to better outcomes.

 Conclusions

While there is significant variability in how hospitalist programs are structured across Canada, our study suggests that it is possible to identify distinct “clusters” of HM programs that are sufficiently distinct from each other. These segments confirm our previous proposed typology and provide an opportunity to identify peer groups and develop operational benchmarks. Benchmarking HM programs may be of significant value for hospital administers who plan to implement new programs, or who are looking to find ways to determine expected outcomes, growth potential and contract assessments, based on peer programs rather than national generalizations.  The program types also demonstrate a significant predictive relationship on a number of outcome metrics such as hospitalist satisfaction with work. Further research is required to uncover any potential associations between program type and clinical outcome measures such as mortality or length of stay.

Acknowledgements

 We like to acknowledge Dr. Scott Evans and Ms. Angie Baker for their contribution to this research and for conducting the statistical analysis. We also like to thank the Canadian Society of Hospital Medicine for giving us access to the survey data.

References

1. Day A, MacMillan L. Neglect of the inpatient: The hospitalist movement in canada responds. Hospital quarterly. 2001;4(4):36. https://www.ncbi.nlm.nih.gov/pubmed/11727487.

2. Yousefi V, Evans M. Hospital medicine maturity framework: Proposing a novel framework for defining stages of hospital medicine in canada. Healthcare Quarterly. 2011;14(4):67-73. http://www.longwoods.com/content/22653

3. Canadian Society of Hospital Medicine. 2012 National Hospital Medicine Survey. Vancouver, BC. 2012

4. White HL, Stukel TA, Wodchis WP, Glazier RH. Defining hospitalist physicians using clinical practice data: A systems-level pilot study of ontario physicians. Open medicine. 2013;7(3):e74. https://www.ncbi.nlm.nih.gov/pubmed/25237402.

5. Cawley P, Deitelzweig S, Flores L, Miller JA, Nelson J, Rissmiller S, Wellikson L, Whitcomb WF, Principles and Characteristics of an HMG. J. Hosp. Med 2014;2;123-128.

6. 6. Manser T. Teamwork and patient safety in dynamic domains of healthcare: A review of the literature. Acta Anaesthesiologica Scandinavica. 2009;53(2):143-151.