Summaries of Research Projects Using the LTC Data Cooperative
Below is a summary of selected projects that have been approved to use the LTC Data Cooperative data. We encourage prospective researchers to refer to this list for examples of research that successfully utilizes LTC Data Cooperative data.
Clinical Trials
Title: TRAIN-AD: Trial to Reduce Antimicrobial Use In NH Residents With Alzheimer's & Dementias
Principal Investigator: Susan Mitchell
Background: The late-stage of Alzheimer’s disease and other dementias is characterized by the onset infections that are widely mismanaged. Antimicrobials are extensively prescribed, most often without evidence to support a bacterial infection. In a prospective study conducted by our group in 2015, antimicrobials were prescribed for 72% of suspected infections of skilled nursing facility (SNF) residents with advanced dementia, but only 44% episodes met guideline-based criteria for treatment. Motivated by these findings, we conducted TRAIN-AD (Trial to Reduce Antimicrobial Use in Skilled nursing facility residents with Alzheimer’s disease and other Dementias) from 2017-21. TRAIN-AD 1.0 was a Stage III (efficacy-effectiveness) trial with mixed traditional and pragmatic trial design features. The multicomponent intervention merged best practices in infectious diseases and palliative care and included provider training and proxy information. Building on this work, the objective is to conduct TRAIN-AD 2.0; a larger, fully pragmatic trial to evaluate the effectiveness of the TRAIN-AD intervention to improve infection management among residents with dementia in 50 SNFs that are members of an integrated provider managed care network (Iowa Health Care Quality Network). The pragmatic design is enabled by leveraging an established Long-Term-Care (LTC) Data Cooperative, an NIH-funded initiative that allows access to the SNFs’ electronic health records for subject characterization and outcome ascertainment.
Study design: Pragmatic clinical trial
Key measures:
-
Compare the number of burdensome interventions (hospital transfers, bladder catheterization, chest x-ray, blood cultures) used to evaluate suspected infections/person-year (secondary outcome) in the intervention vs. control arms among residents with i. moderate to severe dementia and ii. dementia at any stage. Medicare Claims (hospital transfers) and the SNF EHR (other interventions) will be used to ascertain outcomes.
-
Conduct a process evaluation of the TRAIN-AD 2.0 intervention guided by the RE-AIM (Reach, Effectiveness, Adoption, Implementation, and Maintenance) framework. Each RE-AIM component will be assessed using quantitative measures (e.g., participant vs. non-participant SNFs features (Reach), % providers completing training (Implementation)), and qualitative interviews with SNF champions (N=25) and providers (N=25) (e.g., factors enabling Champion engagement (Adoption) and program uptake (Implementation)).
Funding source: NIA – 5R37AG032982-13
Title: Prevention of Injury in Skilled Nursing Facilities through Medication Optimization (PRISM)
Principal Investigators: Cathleen Colon-Emeric, Sarah BerryBackground: Fractures are a leading cause of disability, need for long-term care, and death. Prior studies have shown that we can prevent poor outcomes, including death, by treating patients with osteoporosis medications and stopping risky medications that cause falls and injury. Many patients with a fracture receive care in Skilled Nursing Facilities (SNFs), making SNFs an opportune site to engage patients and their caregivers in medication review. Patients may have a wide range of preferences about whether to prioritize fall and fracture prevention over the management of symptoms like pain, anxiety and depression. Our nurse care managers will provide patients, their families, and providers with information about how much benefit they will get from starting osteoporosis medications or reducing medications that cause falls but may relieve other important symptoms.Study design: Pragmatic clinical trialKey measures: The primary outcome (injurious falls) will be collected using Medicare claims data linked with patient identifiers. As a secondary outcome, we will use the EHR to compare the number of patients treated for osteoporosis and the number of patients on fall-risk increasing drugs at the time of SNF discharge. Other secondary outcomes reflecting patient priorities are validated scales measuring medication burden, falls self-efficacy, pain, anxiety and sleep ascertained during a telephone survey with the patient or their caregiver approximately 90 days after discharge.
Funding source: PCORI - OFP2022C3-30363
Observational Studies
Title: Deprescribing of Diabetes Treatment Regimens in Long Term Care Residents with Alzheimer’s Disease and Related Dementias
Principal Investigator: Medha Munshi
Background: While deprescribing medications with a higher risk of hypoglycemia is now a part of the Standards of Medical Care established by the American Diabetes Association for the general population, implementation of such a strategy for older adults in the nursing home (NH) setting requires incorporating additional geriatric care principles into diabetes management and thus, a distinct skillset and expertise. Many NHs may lack easy access to a specialist or staff (i.e., physicians, nurse practitioners, or physician assistants) with such skills. There is also currently a lack of practical algorithms for deprescribing diabetes treatments for NH residents with Alzheimer’s Disease and Related Dementias (ADRD) and diabetes. These are major barriers to deprescribing and glucose-lowering medication optimization in NHs. Our research team developed and implemented the Simplification of Treatment Regimens and Individualized Diabetes Education (STRIDE) intervention to enhance geriatric diabetes care by educating NH physicians, nurse practitioners, and physician assistants. We provided these clinicians with resources, specifically deprescribing algorithms, to engage in optimization of medications with a higher risk of hypoglycemia among NH residents with ADRD and diabetes. To facilitate a future larger pragmatic trial of the STRIDE intervention, we also placed continuous glucose monitors (CGMs) on NH residents and collected detailed glycemic control data. The CGM data will be used as a gold-standard measure to validate use of medications with a higher risk of hypoglycemia measured routinely in the electronic health records (EHR) as a pragmatic trial outcome (i.e., a proxy for CGM-identified hypoglycemia events).
Study design: Observational comparative effectiveness research (post-intervention)
Key measures: Most critical to this project are: 1) measures of glycemic control like fasting plasma glucose and HbA1c from the LTCDC EHR data; these measures will be used to assess the concordance between the gold-standard CGM measures and the routinely collected EHR data; and 2) measures of medication administration records from the LTCDC EHR data; these measures will be used to assess use of medications with a higher potential to cause hypoglycemia.
Funding source: NIA - IMPACT (U54AG063546)
Title: Using EHR-Medicare linked Data to Examine Responses to and the Impact of COVID-19 in SNFs
Principal Investigators: Vincent MorBackground: The arrival of coronavirus (COVID-19) and the threat that it represents to the population of frail elderly living in skilled nursing facilities (SNFs) is profound. Research done to date on the determinants of a SNF outbreak reveals that the local prevalence of the virus in the community is the most important factor influencing whether residents are infected with COVID-19. A major barrier to developing appropriate clinical and operational responses to the coronavirus pandemic confronting the SNF sector has been the lack of systematic information that is sufficiently current to describe trends and patterns of changes in the incidence rates of COVID-19 diagnosis, the rates of hospital transfer and even mortality rates. The purpose of this study is to document the incidence, prevalence and treatment of COVID-19 as it appeared in US nursing centers by drawing on detailed electronic medical record (EMR) data available from the Long Term Care Data Cooperative made up of EMR data from customers of 3 large EMR vendors and many SNFs from across the country who’ve authorized their EMR vendor to share their own EMR data with the Cooperative.Study design: Observational comparative effectiveness researchKey measures:
-
Defining COVID-19 will be done using a combination of ICD-10 codes or the medication administration record. Some corporations have already created a coronavirus specific user defined assessment in order to identify whether any of their nursing centers has a positive case.
-
Defining Outcomes will focus on hospital transfers, mortality and functional recovery, which come directly from the MDS discharge record and follow-up MDS assessments. We will estimate both the hospitalization and mortality rates of persons diagnosed with COVID-19 as well as for the population of persons with a dementia diagnosis in general since this population may be systematically excluded from hospital treatment due to hospital demand from community cases of coronavirus. Other outcomes ranging from behavioral measures or even vital signs will be explored, particularly temperature increases in which we’ve shown to be strongly predictive of a new COVID-19 infection.
-
Defining Treatments will also be derived from various aspects of the EMR. First, we anticipate a shift in drug regimen (e.g., use of broad-spectrum antibiotics, antiviral therapies, and other treatments) among those with COVID-19 diagnoses which will be identified from the eMAR. Second, we will monitor vital signs such as temperature and blood pressure to determine whether these indicators are predictive of a future diagnosis of COVID-19. We also anticipate working with clinical leadership to characterize facility-level responses to the discovery of an “outbreak” of coronavirus (at least one positive test) qualitatively such that we can create a classification scheme for the initial response: e.g. universal testing, isolation of the person and those with contact with the case, concentration of positive and suspected positive in one wing, etc.
Funding source: NIA – IMPACT (3U54AG063546-02S6, 3U54AG063546-04S1)
Title: Evaluating the Effectiveness of an Algorithm to Identify Important Drug-Drug Interactions that Impact Nursing Home Residents
Principal Investigator: Andrew ZulloBackground: Nursing home (NH) residents need multiple medications for the multiple chronic conditions that they have, which results in polypharmacy. The unintended result of polypharmacy is that it can sometimes increase the risk of harmful drug-drug interactions (DDIs). We must balance the need for medications against the potential risks of DDIs. Many DDI risks for NH residents are unknown, especially for residents living with Alzheimer’s Disease and Related Dementias (ADRD). Information on DDIs has been almost exclusively generated by pharmacology studies in laboratory settings rather than by studies in humans in the real world. To balance the need for medications against the risks of DDIs, we need to identify important DDIs using human data. Ultimately, the information produced by this study will help to refine the DDI warnings in the clinical decision support software that are used by clinicians and improve the outcomes of NH residents.Study design: Observational comparative effectiveness researchKey measures: Based on existing data, we will directly compare the outcomes of residents who experience what appears to be a potential DDI (based on the Long-Term Care Data Cooperative [LTCDC] electronic medication administration records) compared to only one of the two medications or medication classes involved in the potential DDI (i.e., individuals without a potential DDI). We will look at an array of outcomes that are important to NH residents, from changes in physical function to fall-related injuries and hospitalizations. We also will examine whether the effects of DDIs differ for certain groups of people, or subpopulations, who are at higher risk of adverse events (e.g., residents living with ADRD versus those without ADRD).
Funding source: NIA (5R01AG07762003)
Title: Epidemiology and Long-Term Effect of Cardiometabolic Medication Substitution during Post-acute Care
Principal Investigator: Chanmi Park
Background: Each year over 1 million Medicare beneficiaries are discharged to skilled nursing facilities (SNF) from acute care hospitals. Acute hospital-to-SNF transitional period is susceptible to medication discrepancy–inconsistencies in medication regimens across different care sites. Medication discrepancy can take the form of omission, addition, duplication, and substitution. It may arise from miscommunications, inadequate reconciliation processes, and formulary or financial restrictions in hospitals or SNFs. Limited prior research showed that medication discrepancy occurs up to 50% during hospital-to-SNF transitions and typically involves cardiovascular, neuropsychiatric, and anticoagulant agents. Our clinical experience suggests that newer, brand-name, and costly medications, such as those selected as the top 10 expensive drugs in the Medicare Part D spending (e.g., angiotensin receptor neprilysin inhibitor [ARNI], glucagon-like peptide 1 receptor agonists [GLP-1RA], and sodium-glucose cotransporter 2 inhibitors [SGLT2i], are often substituted with less expensive and less efficacious alternatives. This substitution of medications occurs partly because SNFs reimbursement rate is predetermined by the Centers for Medicare & Medicaid Services (CMS) and this rate may not fully cover the expenses incurred when using expensive drugs. To date a systematic examination has not been possible due to lack of a database containing medication administration records in a large number of SNFs. Critical gaps remain as to what proportion of medication substitution is resolved and how medication substitution affects chronic disease outcomes.
Study design: Observational, retrospective cohort study.
Specific aims & key measures:
Aim 1. To determine the frequency and factors associated with medication substitutions of cardiometabolic medications during post-acute SNF stay among older adults after acute hospitalization.
Aim 2. To examine long-term chronic disease outcomes associated with medication substitution during post-acute SNF stay.
Funding source: NIA 1R03AG08873001
Title: Effectiveness of Algorithms to Identify Unintentional Medication Use Cascades among Nursing Home Residents (The “Cascades” Project)
Principal Investigator: Kaley Hayes/Andrew Zullo
Background: Nursing home (NH) residents need multiple medications for the multiple chronic conditions that they have.
Medication use cascades are key drivers of unintentional polypharmacy. A medication use cascade occurs when a new medication is started to treat a side effect from an existing medication. The cascade results in the addition of another medication to manage symptoms, but the cause of the side effect, the original medication, remains. Medication use cascades can occur frequently in the routine clinical care of frail older adults due to their clinical complexity and the common occurrence of new chronic conditions, especially for residents living with Alzheimer’s disease and related dementias (ADRD). There is a critical gap in knowledge regarding the degree of new medication use that results from side effects of existing medications among NH residents. To inform medical providers, consultant pharmacists and long-term care clinicians when additional medications may be the result of medication cascade and could be avoided as well as to reduce provider burden, it is essential to identify true medication use cascades among NH residents.
Study design: Observational, retrospective cohort study.
Specific aims & key measures:
1. Develop algorithms to identify medication use cascades among NH residents, including those with ADRD; and 2. Evaluate the effectiveness of the medication use cascades algorithms using real-world data from the Long-Term Care Data Cooperative by examining whether health outcomes differ between residents with versus without cascades.
We will compare the outcomes of residents who experience what appears to be a potential medication use cascade (based upon the electronic medication administration record) versus residents who receive only the first of the two medications or medication classes involved in the potential cascade (i.e., individuals without a potential cascade who have the first medication that could have resulted in the cascade, but who did not receive the second medication that represents the occurrence of a cascade). We will look at an array of outcomes, from changes in physical function to fall-related injuries and hospitalizations. We also will examine whether the effects of cascades differ for certain subpopulations who are at higher risk of side effects (e.g., residents with ADRD). To examine the cascades detection algorithm among residents with ADRD, we will classify residents as having ADRD or not using a combination of Medicare Claims, Minimum Data Set, and electronic health record (EHR) information on cognitive function.
Funding source: NIA 1RF1AG08954101
Title: Post-Acute Care Medication Use and Functional Recovery in Heart Failure
Principal Investigator: Andrew Zullo / Parag Goyal
Background: After hospitalization for heart failure, many older adults are discharged to skilled nursing facilities (SNFs) to recover and regain independence. However, regaining function and returning home can be especially difficult for individuals with heart failure, partly due to the complex role medications play in recovery. Some medications may promote functional improvement, while others may hinder it. This study seeks to identify which commonly used medications are associated with better functional outcomes and higher likelihood of returning home after SNF care.
Study design: Observational, retrospective cohort study.
Specific aims & key measures:
The study will assess the effects of medication use on outcomes such as physical function, discharge to home, hospital readmission, and mortality. Additional outcomes include participation in rehabilitation and cognitive function during the SNF stay. Findings will help inform more effective medication management to support recovery and safe transitions back to the community.
Funding source: NIA - 1R01AG088522-01
Title: Benefits and Harms of Long-term Osteoporosis Pharmacotherapy: Impact of Treatment Length, Type, Switching, and Holidays
Principal Investigator: Kaley Hayes
Background: More than 90% of female and 70% of male nursing home (NH) residents are affected by osteoporosis. NH residents are at high risk of osteoporotic fractures due to advanced age, high fall risk, multiple comorbidities, use of multiple medications, and impaired physical and cognitive function; and thus indicated for osteoporosis treatment. However, many NH residents are often burdened with multiple other chronic medications and have limited life expectancies, potentially reducing the benefits of osteoporosis treatment. Oral bisphosphonates (i.e., alendronate, risedronate, ibandronate) are first-line therapy for osteoporosis and the most prescribed treatment for osteoporosis. Their effectiveness in reducing incidence of vertebral and nonvertebral osteoporotic fractures has been shown in community-dwelling older adults but not in older adults residing in the NH setting, especially after a baseline period of use. Drug holidays, defined as a temporary discontinuation after an initial BP treatment period, are recommended for some patients to reduce the risk of adverse events associated with long-term bisphosphonate treatment, such as atypical femoral fractures and osteonecrosis of the jaw. There is a critical gap in knowledge regarding the use and effectiveness of long-term osteoporosis medication therapy and drug holidays, in NH residents. Most studies have been among older adults residing in the community who have different risk-benefit profiles than NH residents. It is essential to provide evidence on the effects of long-term osteoporosis treatment strategies that will improve clinical decision making for this vulnerable population. Examining the incidence of fractures and other adverse events among patients with varying risk of fractures during drug holidays can inform who will benefit from BPs. Patients who experience fractures during drug holidays may benefit from continuing BP therapy rather than pausing it.
Study design: Observational, retrospective cohort study.
Specific aims & key measures:
Based on EHR and claims data from residents in nursing homes, we will directly compare the outcomes of residents who continue bisphosphonate treatment, switch to another osteoporosis therapy (based upon the electronic medication administration records and claims), or start a drug holiday (i.e., at least 120 days without bisphosphonate use). We will examine multiple clinical outcomes including fractures and other adverse events (e.g., mortality, hospitalization). We will examine effects in clinically important subgroups (e.g., those with prior fracture). We will then formulate a model to predict fractures after bisphosphonate discontinuation using data from multiple sources, including Medicare claims, Minimum Data Set (MDS), and electronic health records (EHR), which provide information on comorbidities and cognitive function. The algorithm will be designed to use the smallest number of clinical inputs possible to aid NH clinicians in deciding whether to stop bisphosphonate therapy (i.e., initiate a drug holiday) vs. continue treatment. We will then identify a threshold to classify residents as either at high or low risk of fracture during a drug holiday to aid in clinical decision-making.
Funding source: NIA- 5R01AG078759-03
Title: Effectiveness of Stopping Antipsychotic Medication among Individuals Admitted to Nursing Home Facilities
Principal Investigator: Mir M. Ali
Background: Antipsychotic medications are used to treat schizophrenia and other psychotic disorders. Additionally, they may be used as adjunct therapy in mood disorders including bipolar disorder and major depression under certain indications. The benefits of antipsychotic medications can sometimes be outweighed by their side effects and health risks, which can vary from mild to severe. Appropriate assessment for gradual dose reduction and deprescribing are key components of safe prescribing of these medications. It is unclear how many nursing home residents are prescribed antipsychotics prior to nursing home admission, and how frequently these medications are later discontinued. For this observational comparative effectiveness study, we will be using the Long-Term Care Data Collaborative Data linked with CMS Medicare claims data. The objective of this study is to compare outcomes among nursing home residents with a) pre-existing prescription to antipsychotic medication and b) those without a prescription, prior to admission. We will examine the frequency of nursing home clinicians de-prescribing antipsychotic medications for patients after 90 days among both groups to assess potential health benefits and risks of stopping these medications.
Study design: Observational comparative effectiveness research.
Specific aims & key measures:
This will be a comparative cohort study design comparing clinical outcomes in nursing home residents that were exposed to and not exposed to antipsychotics prior to admission (with or without a psychotic condition). To understand de-prescribing frequency, we will compare clinical outcomes and adverse effects (AEs) for patients who a) stopped taking antipsychotic medications within 90 days and those who b) continued antipsychotic treatment, among both groups.
We will measure characteristics of patients and compare AEs among residents with and without a pre-existing prescription. To evaluate the health consequences of deprescribing antipsychotic medication in nursing home residents, we will evaluate antipsychotic drug associated AEs such as neurological effects, stroke, falls, and functional and cognitive impairment as our outcome measures. We will take into account baseline patient, facility level, and geographic characteristics. To account for baseline differences across populations, we will conduct risk adjustments to standardize the comparison groups.
Funding source: NIA (internal funding)
Title: The Impact of Institutional Special Needs Plan Enrollment on End-of-Life Outcomes for Older Adults with Dementia
Principal Investigator: Momotazur Rahman
Background: Skilled nursing facility (SNF) residents with dementia have complex care needs and often experience hospitalizations and other aggressive care at the end of life. The last decade has seen significant growth and diversification in Institutional Special Needs Plans (ISNPs), a type of Medicare Advantage plan for long-stay SNF residents that integrates a capitated payment model with a Medicare-approved model of care focused on care coordination and quality improvement. Most ISNPs employ on-site advanced practice clinicians to manage chronic illness, treat acute conditions in-house, lead advance care planning, and support SNF nursing staff. Additionally, financial quality incentives built into the model reward favorable outcomes. Our preliminary work using national Medicare data has shown that, among long-stay residents with dementia at the end-of-life, ISNP enrollees are significantly less likely to be hospitalized, admitted to an intensive care unit, or intubated in the last month of the life, compared to residents not on ISNP. The ISNP model may have been particularly important for SNF residents with dementia during the COVID-19 pandemic due to their significantly increased risk for COVID-19-related death which prioritized the need for goals-of-care discussions, advance directive completion, and treatment-in-place within the SNF setting. The electronic health record (EHR) data available through the Long-Term Care (LTC) Data Cooperative provide a unique opportunity to integrate advance directives into our existing work.
Study design: Observational, retrospective, longitudinal study.
Specific aims & key measures:
Objectives: To examine the association of ISNP enrollment with advance directive completion and end-of-life outcomes among SNF long-stay residents with dementia who died between 2019 and 2023. We will further assess how these relationships vary over time (i.e. before, during, and after the pandemic), and with respect to periods of high community COVID-19 prevalence during the pandemic.
Outcomes: (1) Advance directive status, measured from EHR orders; (2) End-of-life outcomes, measured from Medicare claims for the last 30 days of life, including: hospitalization, mechanical ventilation, intensive care unit admission, and hospice enrollment
Other Measures: Resident characteristics including date of death, dementia diagnosis, demographic characteristics, and clinical characteristics (including vaccination status) will be measured using a combination of EHR, Minimum Data Set (MDS) and Medicare data. ISNP enrollment will be measured from Medicare enrollment and Medicare Advantage monthly plan data. Facility characteristics such as bed size and zip code will be obtained from CASPER data. Community COVID-19 prevalence will be measured from the Johns Hopkins Coronavirus Resource Center data which is publicly-available.
Funding source: NIA - 1R01AG08209801A1
Title: Integrating Deep Learning and EHR Data to Predict Psychosis, Survival and Fall Outcomes in Nursing Home Residents with Alzheimer's Disease
Principal Investigator: Lirong Wang
Background: Alzheimer's disease and related dementias (ADRD) are common among NH residents, with ~50% developing psychosis, leading to increased care needs, fall-related fractures, weight loss, and skin breakdown. Evidence on treatment effectiveness in reducing psychosis incidence and preventing adverse outcomes remains limited. Our study seeks to evaluate the impact of different treatments on reducing these risks and improving care strategies for NH residents with ADRD. By identifying effective prevention strategies, our findings help nursing home providers optimize care plans to improve residents' quality of life and mitigate adverse health outcomes, including fall-related fractures, unintended weight loss, and skin breakdown.
Study design: Observational, comparative effectiveness research.
Specific aims & key measures:
Aim 1: We will apply DeepBiomarker, a deep learning model, to screen and identify treatments associated with increased psychosis risk. After screening, we will conduct comparative effectiveness analysis using time-to-event methods (Cox models, clinical trial emulations). ADRD will be defined based on both ICD codes and MDS – item I4200. Psychosis will be defined by ICD codes and MDS –item E0100. These definitions would be applied to all aims.
Aim 2: Retrospective cohort study using time-to-event analysis (Cox models, DeepSurv) to evaluate 1-and 3-year mortality. The medications assessed in Aim 2 will include those identified in Aim 1 as increasing the risk of psychosis, as well as other commonly prescribed drug classes in ADRD patients. Subgroup analyses will assess differential medication effects in ADRD residents with vs. without psychosis and across characteristics (age, sex, race, ZIP code, facility type, comorbidities, functional status, etc.). Censoring will occur at death or prior to hospice entry.
Aim 3: Retrospective cohort study evaluating associations between specific medication classes (e.g., antihypertensives, antidiabetics, antipsychotics) and key outcomes -- fall-related fractures, unintended weight loss, skin breakdown and functional decline among NH residents with ADRD. We will then test whether these associations differ between ADRD and psychosis vs. ADRD without psychosis. Time-to-event and mixed-effects models will be applied.
Funding source: NIMH - 2R01MH116046