Multidisciplinary Case Management for Patients at High Risk of Hospitalization: Comparison of Virtual Ward Models in the United Kingdom,United States,and Canada

Abstract

Virtual wards are a model for delivering multidisciplinary case management to people who are at high predicted risk of unplanned acute care hospitalization. First introduced in Croydon, England, in 2006, this concept has since been adopted and adapted by health care organizations in other parts of the United Kingdom and internationally. In this article, the authors review the model of virtual wards as originally described—with its twin pillars of (1) using a predictive model to identify people who are at high risk of future emergency hospitalization, and (2) offering these individuals a period of intensive, multidisciplinary preventive care at home using the systems, staffing, and daily routines of a hospital ward. The authors then describe how virtual wards have been modified and implemented in 6 sites in the United Kingdom, United States, and Canada where they are subject to formal evaluation. Like hospital wards, virtual wards vary in terms of patient selection, ward configuration, staff composition, and ward processes. Policy makers and researchers should be aware of these differences when considering the evaluation results of studies investigating the cost-effectiveness of virtual wards. (Population Health Management 2012;15:315–321)

Introduction

P atients at high risk of unplanned hospital admission typically have complex health care and personal care needs^1,2 and are more likely to live in low-income and underserved areas.^3,4 Such patients account for a large proportion of health care expenditure.⁵ Clinicians and policy makers in many countries are interested in how investments in preventive care might improve the health status and quality of life of these people, while potentially generating net savings to the health care system as a whole from averted acute care hospitalizations.⁶ However, despite the attractiveness of this goal, a number of studies have underlined how difficult it can be to achieve these savings in practice.^7

–10

An important element of any hospital avoidance program is the ability to identify patients at high risk of future unplanned admission to an acute care facility. Unfortunately, doctors, nurses, and case managers do not appear to be able to predict hospital admissions reliably.¹¹ Furthermore, patients who have multiple emergency admissions in a given year tend to have fewer admissions in subsequent years, even without any intervention—a phenomenon known as “regression to the mean.”¹² Therefore, tools that identify patients who currently are experiencing multiple hospital admissions also are unsatisfactory.^13,14 A more reliable way to identify individuals at risk of future unplanned hospital admissions is to use a predictive risk model.^15,16

In many health care systems, predictive models are currently being used to identify which patients should be offered hospital avoidance interventions. Given that the accuracy of predictive models can be quantified, the crucial questions now are whether these interventions are efficacious, equitable, and cost-effective.¹⁷ One such intervention is the virtual ward (VW) (Fig. 1).^18,19 In 2010, Carlisle reported that over 30 sites had established VWs or were in the process of doing so.²⁰ Formal evaluation studies are currently being conducted in 6 of these sites: 4 sites in the United Kingdom, and 1 each in Canada and the United States.^21

–24 However, each of the 6 VW study sites has adapted and implemented the intervention somewhat differently.

FIG. 1.

Original Virtual Wards (VW) model.¹⁸

In this article, the authors first review the VW model as it was originally described, and then outline the ways in which VWs are currently operating in 6 study sites. The authors conclude by exploring a wider range of variations to the VW model that researchers and policy makers should bear in mind when considering this type of population health management.

Description of Care Practice

Virtual wards: Original model

VWs are a model for organizing care in the community for individuals in a population who are predicted to be at very high risk of a future emergency hospitalization.^18,19 They build on a long tradition of hospital-at-home schemes, which may be defined as services that provide, “active treatment by health care professionals, in the patient's home, of a condition that otherwise would require acute hospital in-patient care, always for a limited period.”²⁵

In 2005, a Cochrane review concluded that there was insufficient evidence of so-called “early discharge hospital-at-home” services being cheaper than inpatient care, and recommended that future research should focus on “admission avoidance” hospital-at-home schemes.²⁶ VWs are an example of this second type of hospital-at-home scheme. Rather than waiting for high-risk people to be admitted to an acute care hospital as an emergency, VWs instead bring the coordination of a multidisciplinary team to the patient in advance in order to deliver intensive preventive care in the community.

How are Virtual Wards being implemented in practice?

Following its implementation in Croydon, England, clinicians and health care managers working in other parts of the United Kingdom and overseas began to adopt and modify the VW model in their local areas. Formal evaluations are currently under way in at least 6 of these sites: Croydon, Devon, Wandsworth, and North Somerset in England, as well as in New York City, New York, and Toronto, Ontario. The authors conducted an online survey in April 2011 to ascertain how VWs are currently operating in these 6 study sites. Table 1 presents the findings, organized according to the taxonomy described by Boaden and colleagues, namely: (1) target population; (2) assessment, care planning, and monitoring of patients; and (3) exit.³⁰

Table 1.

Comparison of Selected Virtual Wards

	Description			Target Population		Assessment, Care Planning, and Monitoring of Patients				Exit
Study Site	Project Name	Number of VWs	Number of Patients per Ward	Predictive Model Used	% Patients Identified Using a Predictive Model	Members of ‘Core’ VW Team ^*	Frequency of ‘Core’ Team Meetings ^*	Members of ‘Wider’ VW Team ^**	Frequency of ‘Wider’ Team Meetings ^**	Decision to Discharge
Croydon, England	Virtual community ward (VCW)	8	50	Combined model²⁷	All or almost all	Community matron, ward clerk	Once a day or ad hoc (varies from VW to VW)	District nurses, GP, mental health team	Ad hoc	Clinical decision but also may be guided by changes in risk score. Also, patients may be discharged if they do not engage or if the VCW has nothing more to add to the patients' care.
Devon, England	VW	23	48–150 (some VWs have part-time community matrons and therefore have smaller caseloads)	Combined model²⁷	Between 75% and 100%, depending on the VW	Community matron and complex care team administrator (ward clerk)	Some VWs once a day, others once a week	Adult care services social workers, community psychiatric nurse (CPN), CPN for older people, staff-grade elderly care doctor, physical therapist, occupational therapist, voluntary sector representative, district nurses. In addition, some VWs have an associated GP, social care manager, and complex care team manager	Weekly for some VWs; for others about once a month	Combination of clinical decision (about when admission objectives are achieved) and risk score (ie, patients discharged when score falls)
Wandsworth, England	Wandsworth VW project or community VW	4	40–50	PARR model⁴⁰	About 25%	Virtual ward GP (VWGP), community matron, ward clerk	Once a day	Social worker, district nurse, physical therapist, occupational therapist, pharmacist, drug & alcohol therapist. Plus access to specialist nurses (heart failure, diabetes, palliative care), social worker, community mental health team, and a dietitian	Weekly or monthly, depending on the VW	Clinical decision
North Somerset, England	Community ward	7	35	None	None	Community matron, district nurse, occupational therapist, physical therapist, team coordinator, team administrator.	Once a day	Pharmacists, social worker, community development worker physical therapist, occupational therapist	Once a day	14-day fixed length of stay on VW
Toronto, Ontario	Virtual ward	1	No maximum set yet, typically about 45	LACE model³¹	All (or almost all)	One physician, 2 care coordinators and a team assistant (ward clerk)	Once a day	Pharmacist, nurse practitioner, mental health case manager	About 2 or 3 times a week	Clinical decision
New York City, New York	Hospital2Home	3	200 (1 ward) or 100 (2 wards)	Modification of a model described in Health Affairs, which uses New York State Medicaid data³²	All (or almost all)	Community-based care managers, social work supervisors, housing coordinator, program manager, and an administrative assistant	About once a week	Primary care physician; psychiatry fellow	About once a week	Do not generally discharge patients unless mandated by New York State Department of Health

Information was collected in April 2011 using an online survey tool.

Members of the “core team” spend most or all of their time working with VW patients.

Members of the “wider team” spend less than half of their time working with VW patients.

GP, general practitioner; VW, virtual ward.

As can be seen in Table 1, the original VW model has been modified in a number of ways by the different study sites. For example, in Wandsworth, England, 4 VWs have been established, each led by a general practitioner, known as a “VWGP,” who works full time exclusively with VW patients (Fig. 2).³³ In New York City, the “Hospital2Home” project shares many common features with VWs but, to reflect the high prevalence of homelessness among high-risk patients, the multidisciplinary team includes housing officers, and patients are issued pre-programmed mobile phones to ensure that they can contact the team.³⁴ Finally, in the Toronto VW project, patients are identified using a predictive model called LACE, which forecasts risk of readmission within 30 days of discharge from hospital.³¹ High-risk patients are offered care by a VW team that is led by a hospital physician and includes a range of community-based professionals, which may help improve integration between hospital and community services.³⁵

FIG. 2.

Example case history (certain details have been changed to protect anonymity).

Variations of the Virtual Ward model

Our survey and visits to the study sites have revealed a range of factors that determine how VWs have been implemented in practice. Table 2 provides an outline of a more detailed framework that researchers and policy makers might wish to consider when interpreting the various forthcoming VW evaluation findings.

Table 2.

Elements of Virtual Wards That May Vary Locally

Domain	Details of How Virtual Ward Is Organized
Patient selection	• Whether a predictive model is used to select all, some, or no VW patients • Which predictive model is used • Whether an “impactibility model” is used to identify the subgroup of high-risk patients believed to be most amenable to preventive care³⁶
Configuration	• Whether VWs are organized around groups of GP practices or equivalent • Whether VW catchment areas reflect the distribution of high-risk patients living in a local area.
Staff	• Whether the VW staff mix reflects the needs of local high-risk patients • Whether the clinical team includes a ward clerk to act as a central point of contact for coordinating the VW • What arrangements are in place to provide care after hours
VW processes	• Whether new VW patients are screened for particular conditions (eg, depression, cognitive impairment) • Whether a single electronic medical record is used that enables VW staff to read and enter patient notes, and whether this record can be accessed remotely. • Are alert systems in place to inform VW staff if patients have any contact with urgent care providers, including local emergency departments? • How frequently do ward rounds take place, and can VW staff join the meeting by teleconference? • What is the policy for discharging VW patients (eg, fixed length of stay, clinical decision, guided by reduction in predictive risk score). • Whether the VW staff has been offered additional training in hospital avoidance techniques such as motivational interviewing. • Whether techniques such as Lean have been used to optimize the efficiency of VW processes³⁷
Evaluation and monitoring	• Are “Key Performance Indicators” used to monitor such factors as staff attendance at ward rounds and the turnover of VW patients? • Has any evaluation considered regression to the mean? Simple pre/post evaluations may be misleading. Better alternatives are randomized controlled trials, prognostic score matching, and regression discontinuity analysis. • Are local hospitals responding to any reduction in hospital admissions by VW patients by admitting lower risk patients or by increasing the length of stay for hospital inpatients (known as “Roemer's Law”)?³⁸

GP, general practitioner; VW, virtual ward.

Discussion

VWs bring together predictive modeling and a hospital-at-home approach with the aim of improving the quality and integration of preventive care in the community for patients at high risk for unplanned hospital admission. This model of preventive, integrated care may have a number of theoretical advantages over “traditional” one-to-one case management. Potential advantages include the ability to provide continuity of care when a member of the team is away sick or on annual leave, and to reduce fragmentation between the primary, community, and social care sectors by bringing together professionals from different organizations to participate in a joint “ward round.” However, the efficacy and cost-effectiveness of VWs are currently unknown. Several evaluations of VWs are under way and their results are awaited; this article highlights the considerable variation that exists between different VW study sites in practice.

In our survey, some study sites reported that they are using the term “virtual ward” or similar, but currently are delivering care that differs from the core definition of a virtual ward—either because they are not using a predictive model to identify patients (eg, North Somerset) or because they are not holding regular multidisciplinary “ward rounds” (eg, Croydon). Equally, another site (the Hospital2Home project in New York City) appears to be providing the core elements of the VW model of care without using the term “virtual ward.” Therefore, it will be important for researchers and policy makers to bear these differences in mind when the findings of the evaluations are published.

Finally, it is important to remember that even if the evaluations do show a comparative reduction in unplanned hospital admissions for VW patients, any savings generated by the intervention could be undermined by Roemer's law, which states that “a hospital bed built is a hospital bed filled.”³⁸ In other words, local hospitals may be expected to respond to any reduction in risk for those patients receiving the VW intervention by admitting other, lower risk patients and/or by increasing the average length of stay for hospital inpatients. It will be important for researchers and commissioners to monitor carefully for this phenomenon. One rational way to counter Roemer's law is to close hospital beds or hospital wards. However, politicians and the general public tend to have strong emotional attachments to hospitals, partly because issues of access are more observable than quality or cost-effectiveness concerns. As a result, ward closures are notoriously difficult to achieve in practice, even where they are preferable, all things considered.³⁹ One unintended benefit of VWs, therefore, is that they could conceivably make the closing of a hospital ward slightly less unpalatable to the public if the closure of each hospital ward was coupled with the opening of 1 or more VWs.

Conclusion

VWs are a model of multidisciplinary case management that is being adopted and adapted in many areas, and which is currently being evaluated in several studies. However, in the same way that different hospital wards vary from each other in terms of their staffing, configuration, and daily routines, so too a number of different variants of VWs are emerging. Researchers should be mindful of these differences when interpreting the results of evaluation studies of VWs.

Footnotes

Acknowledgment

The authors are grateful to Martin Bardsley, Ph.D., who provided helpful comments on an earlier draft; to Heather Watson, M.D., who provided the case history; and to the representatives from the 6 study sites who completed the online questionnaire.

Disclosure Statement

Dr. Lewis, Mrs. Wright, and Dr. Vaithianathan disclosed no conflicts of interest.

This work was funded by the National Institute for Health Research (NIHR) Service Delivery and Organization (SDO) program (Project number 09/1816/1021). The views and opinions expressed here are those of the authors and do not necessarily reflect those of the NIHR SDO program or the Department of Health.

References

Hastings

, Horney

, Landerman

, Sanders

, Hocker

, Schmader

. Exploring patterns of health service use in older emergency department patients. Acad Emerg Med, 2010; 17:1086–1092.

Mudge

, Kasper

, Clair

et al. Recurrent readmissions in medical patients: A prospective study. J Hosp Med, 2011; 6:61–67.

Blunt

, Bardsley

, Dixon

. Trends in Emergency Admissions in England 2004–2009. London: Nuffield Trust, 2010. http://www.nuffieldtrust.org.uk/sites/files/nuffield/Trends_in_emergency_admissions_REPORT.pdf. 2011 September 19.

Raven

, Billings

, Goldfrank

, Manheimer

, Gourevitch

. Medicaid patients at high risk for frequent hospital admission: Real-time identification and remediable risks. J Urban Health, 2009; 86:230–241.

Raising the Profile of Long Term Conditions Care: A Compendium of Information. London: Department of Health, 2008.

Billings

, Mijanovich

. Improving the management of care for high-cost Medicaid patients. Health Aff (Millwood), 2007; 26:1643–1654.

Gravelle

, Dusheiko

, Sheaff

et al. Impact of case management (Evercare) on frail elderly patients: Controlled before and after analysis of quantitative outcome data. BMJ, 2007; 334:31.

Steventon

, Bardsley

, Billings

, Lewis

. An Evaluation of the Impact of Community-Based Interventions on Hospital Use: A Case Study of Eight Partnership for Older People Projects (POPP) London: The Nuffield Trust, 2011.

Centers for Medicare and Medicaid Services. Medicare Coordinated Care Demonstration. http://www.cms.hhs.gov/DemoProjectsEvalRpts/downloads/CC_Fact_Sheet.pdf. 2009 November 7.

10.

Purdy

. Avoiding Hospital Admissions: What Does the Research Evidence Say? London: The King's Fund, 2010.

11.

Allaudeen

, Schnipper

, Orav

, Wachter

, Vidyarthi

. Inability of providers to predict unplanned readmissions. J Gen Intern Med, 2011; 26:771–776.

12.

Roland

, Dusheiko

, Gravelle

, Parker

. Follow-up of people aged 65 and over with a history of emergency admissions: Analysis of routine admission data. BMJ, 2005; 330:289–292.

13.

Duncan

. Healthcare Risk Adjustment and Predictive Modeling. Winstead, CT: ACTEX Publications, Inc., 2011.

14.

Curry

, Billings

, Darin

, Dixon

, Williams

, Wennberg

. Predictive Risk Project: Literature Review. London: The King's Fund, 2005.

15.

Cousins

, Shickle

, Bander

. An introduction to predictive modeling for disease management risk stratification. Dis Manag, 2002; 5:157–167.

16.

Georghiou

, Blunt

, Steventon

, Lewis

, Billings

, Bardsley

. Predictive Risk and Health Care: An Overview. London: Nuffield Trust, 2011. http://www.nuffieldtrust.org.uk/sites/files/nuffield/publication/Predictive-risk-and-health-care-an-overview_0.pdf. 2011 September 19.

17.

Ham

. Chronic care in the English National Health Service: Progress and challenges. Health Aff (Millwood), 2009; 28:190–201.

18.

Lewis

. Croydon Case Study. Case study: Virtual wards at Croydon Primary Care Trust. www.kingsfund.org.uk/document.rm?id=6746. 2011 November 15.

19.

Lewis

. Predictive Modeling in Action: How “Virtual Wards” Help High-Risk Patients Receive Hospital Care at Home. New York: The Commonwealth Fund, 2010.

20.

Carlisle

. How to use predictive risk modeling in healthcare. Health Service Journal, 25 May 2010. http://www.hsj.co.uk/resource-centre/best-practice/how-to-use-predictive-risk-modelling-in-healthcare/5014244.article. 2011 September 19.

21.

Lewis

, Bardlsley

, Vaithianathan

et al. Do ‘virtual wards’ reduce rates of unplanned hospital admissions and at what cost? A research protocol using propensity matched controls. Int J Integr Care, 2011; 11:e079Epub Jun 30, 2011.

22.

Steventon

, Lewis

, Bardsley

. Evaluation of the North Somerset Community Wards. http://www.nuffieldtrust.org.uk/our-work/projects/evaluation-north-somerset-community-wards. 2011 September 19.

23.

Clinical Trials.gov A virtual ward to reduce readmissions after hospital discharge. http://clinicaltrials.gov/ct2/show/NCT01108172. 2011 September 19.

24.

New York State Department of Health. Chronic Illness Demonstration Projects: Request for Proposals. http://www.health.ny.gov/funding/rfp/inactive/0801031003/0801031003.pdf. 2011 September 19.

25.

Shepperd

, Iliffe

. Hospital at home versus in-patient hospital care. Cochrane Database Syst Rev, 2000; 2:CD000356.

26.

Shepperd

, Iliffe

. Hospital at home versus in-patient hospital care. Cochrane Database of Syst Rev, 2005; 3:CD000356.

27.

Wennberg

, Siegel

, Darin

et al. Combined predictive model: Final report and technical documentation. www.kingsfund.org.uk/document.rm?id=6745. 2011 November 15.

28.

Department of Health. The NHS Improvement Plan. London: The Stationery Office, 2004.

29.

Spending Review. Public Service Agreements 2005–2008. London: HM Treasury, 2004.

30.

Boaden

, Dusheiko

, Gravelle

et al. Evercare evaluation: Final report. Manchester, UK: National Primary Care Development Centre, 2006. http://www.medicine.manchester.ac.uk/primarycare/npcrdc-archive/Publications/NOV06%20EVERCARE%20FINAL%20REPORT.pdf. 2011 November 16.

31.

van Walraven

, Dhalla

, Bell

et al. Derivation and validation of an index to predict early death or unplanned readmission after discharge from hospital to the community. CMAJ, 2010; 182:551–557.

32.

Billings

, Mijanovich

. Improving the management of care for high-cost Medicaid patients. Health Aff (Millwood), 2007; 26:1643–1654.

33.

NHS Wandsworth. Community Virtual Wards. http://www.wandsworth.nhs.uk/gettherighttreatment/Communityvirtualwards/Pages/home.aspx. 2011 November 16.

34.

New York City Health and Hospitals Corporation. Frequent ER users are tracked and connected to services. http://www.nyc.gov/html/hhc/html/newsletter/201007-hospital-2-home.shtml. 2011 November 16.

35.

Lewis

. Toronto virtual ward: Useful lessons for NHS hospitals? http://www.nuffieldtrust.org.uk/blog/toronto-virtual-ward-useful-lessons-nhs-hospitals. 2011 September 19.

36.

Lewis

. “Impactibility models”: Identifying the subgroup of high-risk patients most amenable to hospital-avoidance programs. Milbank Q, 2010; 88:240–255.

37.

Kim

, Spahlinger

, Kin

, Billi

. Lean health care: What can hospitals learn from a world-class automaker? J Hosp Med, 2006; 1:191–199.

38.

Ginsburg

, Koretz

. Bed availability and hospital utilization: Estimates of the “Roemer effect.” Health Care Financ Rev, 1983; 5:87–92.

39.

Vaithianathan

, Lewis

. Operational independence for the NHS. BMJ, 2008; 337:a497.

40.

Billings

, Dixon

, Mijanovich

, Wennberg

. Case finding for patients at risk of readmission to hospital: Development of algorithm to identify high risk patients. BMJ, 2006; 333:327.