Multidisciplinary Mobile Unit for Preventing Hospitalization of Nursing Home Residents

Overview

Elderly patients residing in nursing homes are particularly at risk of experiencing urgent medical problems needing admission to the Emergency Department (ED). This circumstance contributes to ED overcrowding, increases the risk of ward admission of elderly patients, and puts them at an even higher risk of hospitalization-related adverse events. The study hypothesis is that a complex intervention, delivered directly in nursing homes by hospital physicians in case of urgent medical problems, would contribute to reduce hospitalization of older nursing home residents. The intervention consists in a hospital-based "multidisciplinary mobile unit" (MMU), composed of a hospital specialist and a resident in emergency-urgency medicine who are coordinated by a senior physician serving as "flow manager". The team is active on work days, 8 am to 6 pm, and is activated by general practitioners of nursing homes, in case of urgent medical needs of one of the residents. The activation is made by a phone call to the "flow manager", who triages the clinical needs of the case. The output of the phone consultation may include therapeutic advice provided by phone, immediate on-site visit by the MMU team (specialist and resident), scheduled visit by the MMU team, or direct admission to the hospital unit where MMU is based, avoiding ED visits. The MMU team is provided with a portable ultrasound system, an essential set of drugs and medical devices useful in a urgency setting (central venous lines, nasogastric tubes, rectal tubes, bladder catheters). During on-site visits, the MMU team performs diagnosis, stabilization and therapeutic advice, with the mission of avoiding ED visits and hospital admissions whenever possible. The MMU intervention is already active in two nursing homes, since December 2018. The aim of this prospective, pragmatic, multicenter, quasi-experimental study (sequential design with two cohorts) is to test the effects of the implementation of the MMU care model in terms of reduction of unplanned hospitalization rates (primary outcomes), mortality, health service use and costs (secondary outcomes). Two nursing homes (i.e., the ones who already benefit from the intervention) will serve as study group, and two nursing homes with similar geographical location will serve as control group. All residents of the participating nursing homes will be eligible for study inclusion. The study will last for 18 months, and a number of 338 residents is planned for inclusion.

The study is based in the University Hospital of Parma, which has a catchment area of more than 400,000 inhabitants, of whom 22.3% is over 65 years old. It provides the only Emergency service of the district, and it ranks fourth in Italy by number of ED visits (yearly average of over 110,000). The average admission rate of the adult ED population is 18%, of which 65% concern people older than 65. In the last two decades, the University Hospital of Parma has implemented several innovative initiatives to manage the hospital flow of frail multimorbid patients and their complex needs. These initiatives included bed management to avoid "bed-blockers", physician accountability for the discharge process, and creation of a dedicated hospital unit, organized by intensity of care to anticipate the needs of these patients preserving high performance indices. This unit, called Internal Medicine and Critical Subacute Care Unit, performs over 3,500 urgent admissions of frail multimorbid elderly patients per year, with an average length of stay that in 30% of cases is lower than 3 days. Participating nursing homes are public facilities which ensure the presence of nursing staff 24 hours a day and of a physician at least 4 hours a day (high-intensity care facilities). The possible role of distance to the hospital is considered by including in each group one nursing home located next to the hospital and one located \>5 km of distance. This study follows a multimethod approach, based on the MRC framework for developing and evaluating complex interventions, including the development, feasibility assessment, and evaluation phases. 1. Development of the intervention First, the different types of approaches reported in the literature, described above, were considered. The "prevention approach", interventions conducted in nursing homes, was chosen as the most suitable strategy to integrate the hospital's organizational model already in place, as it can target both hospitalization rates and ED overcrowding, allowing to intervene before the person accesses the hospital. Available evidence also prompted us to opt for a multicomponent approach. In fact, data from qualitative interviews reveal that the decision to transfer residents to hospital may be influenced by different factors, such as staffing and skill mix in the nursing homes, treatment options available in the facility, end-of-life decision-making, and communication and bureaucratic requirements. This multifactorial association means that a multicomponent intervention is likely to be more effective than a single-component intervention. The choice of employing a mobile geriatric specialist service was supported by the positive results obtained by the two controlled studies which examined similar interventions. One study evaluated a service where a physician did regular and on-call visits intended to provide services otherwise associated with hospitalization. The other study evaluated an ambulant team with a nurse and a physician, doing comprehensive geriatric assessments of residents as well as reviewing medications and providing support to staff. The proposed intervention does not involve a nurse, unlike the second of the mentioned studies, because in the participating facilities nursing staff is available 24 hours a day. Unlike the experience of the second study, moreover, periodic visits on site will not be performed in the present investigation, since routine clinical management and scheduled follow-up is already performed by nursing home physicians. Finally, medical hospital staff was preferred to community geriatricians, on the assumption that older patients may feel more comfortable being handled by physicians who may have already cared for them at the hospital. Moreover, hospital staff enables direct patient referral to the ward. Finally, this allows the use of diagnostic technologies available at the hospital, which can be used immediately without the need for hospital admission. The MMU care model intervention The model hinges on the strong collaboration between hospital and nursing home staff to provide residents with patient-centered care. It entails a multicomponent intervention which is integrated in standard care and comprises three steps: 1) MMU team activation, 2) on site visit by a team of physicians with geriatric expertise, 3) interdisciplinary care planning. Step 1: MMU team activation Patient selection is necessary to ensure that available resources are used for patients who may really benefit. To this end, the nursing home physician contacts by phone the "flow manager", a skilled internist with strong clinical expertise, organizational attitude and managerial training, during the 8 a.m.-6 p.m. time frame, Monday to Friday. The phone consultation is reported on a form containing the description of the patient's clinical condition and a summary of the conversation. The form also indicates which decision was reached among the following 6 not mutually exclusive options: 1. The patient can be managed by nursing home staff, therapeutic advice is provided by phone 2. Remote reassessment is scheduled after a number of hours agreed upon by the team 3. The MMU team is dispatched for evaluation, treatment and stabilization on site 4. A significant change in vital parameters is observed which requires immediate activation of emergency services 5. Direct hospital admission is considered necessary 6. Ambulatory outpatient visits or tests are planned Step 2: on site visit by a team of physicians with geriatric expertise Visits at the nursing home are performed by two members of the MMU team: an expert hospital physician chosen on a case-by-case basis among the clinical staff of the Internal Medicine and Critical Subacute Care Unit, which comprises internists, gastroenterologists, geriatricians, specialists in clinical nutrition, depending on the disease or clinical problem that must be treated, and a specifically trained resident in Emergency Medicine. The team is provided with a car to reach the nursing homes, a portable ultrasound system, and an essential set of drugs and medical devices useful in an emergency setting. The ultrasound system is equipped with three probes (convex, linear, and phased-array) for performing thoraco-pulmonary, cardiac, vascular, abdominal and soft tissue ultrasound, when required. Available drugs include those that can be administered intravenously for treating urgent conditions (e.g. loop diuretics, steroids, fluids, antibiotics). Devices include central and peripheral venous lines, naso-gastric and rectal tubes and bladder catheters. Blood tests can also be performed. Step 3: interdisciplinary care planning. Based on the results of the visit and of any performed investigations, the MMU team formulates personalized advice and referrals, and discusses these with the nursing home physician. If stabilization on site is not deemed possible, the MMU team plans a direct admission to the Internal Medicine and Critical Subacute Care Unit, thus avoiding ED access. The planning and the final outcome of the intervention are recorded in the second part of the form. 2. Feasibility assessment A pilot phase of 5 months (December 2018-April 2019) was conducted in two nursing homes in order to look at feasibility of the MMU care Model described above. Before the intervention was introduced, meetings were held with nursing home staff to agree on activation modalities. In this period, 99 phone calls were received, of which 84 required MMU team onsite visits, and 15 were managed with remote consultancy. Of the latter, 3 required direct admission after remote phone consultancy. Only 4 of the 84 patients visited onsite required direct admission. One patient was sent to the ED for massive intestinal bleeding. This phase demonstrated the feasibility of the intervention, and did not highlight any need for modifications. 3. Evaluation phase Aim and objectives The study aim is to verify the effects of the implementation of the MMU care model tested in the pilot phase. Primary objective is to verify reduction of unplanned hospitalization rates in the nursing homes of the intervention group compared to the nursing homes in the control group. Secondary objectives are to measure the effects of the intervention in terms of mortality, health service use, and costs. Study Design. This study is a prospective, pragmatic, multicenter, quasi-experimental study (sequential design with two cohorts), in which usual nursing home care is compared to care provided by applying the MMU model. Study Population. All residents of the participating nursing homes are eligible, regardless of their clinical status. Residents who do not provide informed consent will be excluded. Usual Care. Patients in the control cohort receive usual care, which means the actions to take are decided by the nursing home staff. Generally, this implies that patients who are clinically unstable, or require urgent instrumental tests, will be sent to the ED. Measures. Baseline variables. Demographic data on gender and age are collected by chart review. Outcome variables. The primary outcome is hospitalization rate, considering at the numerator all unplanned admissions occurred during a 1-year period, and at the denominator the sum of the person-time of the at risk population (days of stay at the nursing home). For the intervention group, the numerator corresponds to options c) and d) defined in "Step 1: MMU team activation". Secondary outcomes: * Crude all-cause Death Rate (CDR): the number of deaths during a 1-year period on person-time of the at risk population * Hospital Mortality rate: the frequency of patients who die while in the hospital (death rate/1000) * Length of stay (LOS): the duration of a single episode of hospitalization. Inpatient days are calculated by subtracting day of admission from day of discharge. * Adverse events or complications: frequency of events occurred within 48 hours from MMU team activation and subsequent patient stabilization, for which hospital access becomes necessary. * Costs analysis, comparing the cost differences in the two groups Data Collection. Patient demographic and clinical characteristics are collected at baseline to describe the study population and determine factors associated with hospital rate. Participants' files and electronic data are stored securely at the study site (e.g. locked area, password protected hard- and software). Data integrity will be scrutinized with several strategies (e.g. valid values, range checks, consistency checks). Patient data are only identifiable with the unique participant's number. Personal information will be collected and saved in a separate file (on a different server) which can only be accessed by the Principal Investigator (PI). This information will be used by the PI to retrieve data on any hospital admissions (length of stay, in-hospital death …) from administrative databases (discharge summaries, ED data, Death Registry). Residents' identification data will be deleted once the study is completed, making the dataset anonymous. All study protocol authors will have access to the anonymous dataset. Cost analysis. The investigators will identify the changes in net costs associated with one-year exposure to the intervention, consisting in the induced costs due to incremental resource inputs for carrying out the intervention and hospital health service utilization costs. Staffing costs will be calculated considering the time spent by the professionals involved in the intervention. Non-staff running costs include expenses of MMU staff travelling to and from the nursing home. The health service utilization costs will be identified based on the Diagnosis Related Group (DRG) system. Statistical Methodology. Sample size calculation. The number of subjects to include was estimated using the findings of one study, a large quasi-experimental trial. Thus, considering 56 cases vs 32 cases per 100 residents, and using a 2-sided, large-samples z-test of the Poisson incidence rate difference at a significance level of 0.05, and with a power of 0.90, overall 338 residents should be enrolled. Statistical analysis. Descriptive statistics will be used to summarize patient populations and will be presented as means and standard deviations (SD) when normally distributed, or as medians and interquartile ranges (IQR). For the primary analysis the investigators will used Poisson regression with robust standard errors (SEs) to evaluate relative differences in hospital rates among the two cohorts while adjusting for demographic characteristics. Concerning the secondary outcomes, the following analyses will be performed: * Rates will be compared considering the quotient between the intervention and control groups * A lognormal model will be used to compare in-hospital LOS. * Chi square tests will be conducted for categorical data as adverse events or complications * For costs, the investigators will use the following equations to summarize the annual net costs associated with the implementation of the intervention. Any costs with negative values mean "savings" and any costs with positive values mean "losses". Net costs ˆ A…(intervention costs) +‡ B (Costs for differences in hospital health service utilization) where: A= intervention: staffing costs+intervention: non- staff costs and B= Costs for differences in inpatient care utilization. Therefore, the net costs arising from one-year implementation of the intervention as compared with the current practice will be obtained, where a negative value of net costs represents "cost-saving" and a positive value represents "not cost-saving" The demographic and clinical variables which influence the outcome with a p value\<0.20 in the univariate analysis will be included in the Poisson regression model. The analyses will be performed using SAS 8.2 (SAS Institute, Cary, NC, USA) and STATA-SE 11 (Stata Corp LP, College Station, TX, USA).