Long Term Services and Support Model Components

Published 2017

In this module:

This module describes the data, assumptions, and methods used to adapt HWSM to model the sector-specific long term services and support (LTSS) workforce. LTSS includes nursing homes, residential care facilities, home health, hospice, and adult day services centers. Because of data limitations, home health and home-based hospice visits were combined into home care. MEPS data does not distinguish between home health visits associated with chronic care management and visits following hospital discharge for acute conditions.

Modeling Supply

HWSM supply projections focus on occupations with high education requirements. These requirements create time lags to train new workers. Information on future adequacy of supply for these workers can help mitigate supply inadequacies. Such occupations usually require a license. Licensing databases often can provide estimates of the current year supply. Licensure data are unavailable for direct care workers. For many health occupations, there is no centralized location to obtain licensure data. Rather, such data would need to be obtained from individual state licensing boards. Therefore, the 2015 American Community Survey is the source for much of the workforce supply data used for the LTSS workforce analysis (Exhibit 38).

The main strengths of the ACS are:

  • the availability of occupation code and industry code identifying LTSS setting
  • data are collected by the U.S. Census Bureau for a large sample of the population in each state
  • data are collected annually
  • there is a wealth of information collected on labor force participation, hours worked, and characteristics of workers
    • including demographics and education level.

There are, however, limitations with using ACS to analyze the LTSS workforce:

  • Nurse aides, home health aides and psychiatric aides are aggregated in the ACS data into one occupation comprising all aides. Therefore, we supplemented the ACS data with OES data to estimate the portion of aides that were nurse aides, home health aides, and psychiatric aides (Exhibit 39). However, for modeling, we categorize all home health aides under the home health setting.
  • Some occupation-industry combinations reported in ACS can be unclear. For example, a home health agency owned by a hospital might be categorized under “hospital” for industry.

Supply modeling for several occupations that work in LTSS settings is described elsewhere in this report. This includes RNs and LPNs (discussed in The Nursing Model Components); behavioral health providers (discussed in Behavioral Health Care Provider Model Components); and physicians, APNs and PAs (discussed in Specialist Physician, Advanced Practice Nurse, and Physician Assistant Model Components).

While demand for these occupations is modeled by care delivery setting, supply is not. Comparisons of supply and demand for these occupations could help inform overall adequacy of future supply. This could highlight implications for LTSS, which tends to pay lower compensation relative to acute care settings that might employ these health professionals. Say the overall supply of nurses is projected to be more than adequate to meet demand for services across the health care sector. Then within a particular employment sector such as nursing homes, there is a greater likelihood that supply will be adequate. This is compared to a situation where there were projections of a system-wide occupation shortage.

Modeling future supply of direct care workers is challenging for the following reasons:

  • There are low barriers to entry into the occupation. States have either no formal training requirements or minimal requirements. Hence, there is little information on the numbers of people entering this occupation each year.
  • There are few barriers to leaving the profession. Occupations such as physicians have lengthy and expensive training that reduces the likelihood a person will change occupations. Aides, however, are likely sensitive to earnings. They can move in and out of the direct care workforce based on how earnings as a direct care worker compare to earnings from other occupations. Hence, the direct care workforce experiences high rates of turnover. Previous research suggests that wages of direct care workers are most sensitive to higher minimum wage, lower unemployment rate, and higher rate of overall Medicaid long-term care spending.1

Analysis of ACS provides some insights on the potential future size of aide supply. Direct care workers are disproportionately female and minority (Exhibit 40). There are an estimated 2.6 million individuals working as a direct care worker in a LTSS role. This is equivalent to approximately 2.3 million FTEs (reflecting that some work part time). Together, these FTEs represent 1.4% of the employed workforce in the U.S. in 2015. Only 0.2% of employed white, non-Hispanic males worked as a LTSS direct care worker. Approximately 6.2% of black females worked as a direct care worker.

Populations with greater propensity to be direct care workers (Hispanics, blacks) are growing more rapidly than populations with lower propensity to be direct care workers (non-Hispanic white and other races). If the propensity to be a direct care worker within each demographic group were unchanged, this suggests about 14% growth in direct care workforce supply between 2015 and 2030. This would be an increase from 2.3 million FTEs to over 2.6 million FTEs. Such supply growth likely will be insufficient to keep up with projected growth in demand for services. There is, however, great potential to rapidly grow the direct care workforce simply by increasing wages.

Modeling Demand

The projected demand for LTSS and workforce is derived from the common model estimated on the baseline population and health care usage as outlined in Modeling Demand for Health Care Services and Providers. HWSM already models the demand of many occupations relevant to LTSS (e.g., RNs, LPNs, nurse and home health aides). These projections have been refined for modeling LTSS settings. Previous efforts to model LTSS settings used simplifying assumptions. These included modeling growth in demand for nursing homes and residential care services strictly as a function of an aging population (specifically, the population age 75 and older). Areas of enhancement to demand modeling include:

  • refining the relationship between patient characteristics and economic factors and use of LTSS services
  • adding the adult day care setting
  • including estimates for unpaid care demand, adding occupations to the model
  • refining the scenarios modeled (considering possible changes in care use and delivery patterns).

The population file used for modeling demand was updated to include representative samples of the community-based, residential care-based and nursing home-based populations as noted in Modeling Demand for Health Care Services and Providers. Historically, a matching algorithm was used to combine the latest data from ACS, BRFSS, and NNHS to construct the population file. Starting with this analysis of the LTSS workforce, a representative sample of the population residing in a residential care facility was added. Previously, this population was modeled as living in the community. We identified beneficiaries in the MCBS who reside in a residential care facility. We then used this sample to construct a representative sample of the population in each state living in residential care. Likewise, we used CMS’s 2015 Nursing Home MDS to develop a representative sample of the population in nursing homes.

Projected baseline demand for LTSS assumes that recent patterns of care use and delivery would remain unchanged within each demographic group. Demographic groups is defined by age, sex, and race-ethnicity. Predicted probabilities were applied to the simulated micro-data set for future years. This was done to obtain projected service use specific to the settings that employ long-term care occupations.

For modeling demand for adult day service center, probabilities were assigned to specific population cohorts defined by age group. These probabilities were applied to the population database. The target population is identified as people living in communities with any cognitive difficulty. These probabilities, based on age-distribution of adult day service center patients, were obtained from the National Study of Long-Term Care Providers.2  Approximately 4,800 adult day service centers reported employing around 23,100 FTE nurses and social workers.3  The workforce modeled for the LTSS projections include an estimated 13,700 nurse aides, 4,100 RNs, and 2,500 LPNs working in adult day service centers.

The LTSS component of HWSM currently models demand for approximately 30 professions defined by occupation and medical specialty (for physicians). Many of these professions employ few workers in LTSS. HWSM used provider staffing patterns to convert demand for LTSS into demand for the relevant occupations. These staffing patterns were applied to the constructed population database. Baseline state and national projections were then generated by LTSS setting and occupation. We divided the workload driver for each setting by estimates of FTE providers (Exhibit 38) from the 2015 ACS. This produced staffing ratios for home health, nursing homes and residential care facilities (Exhibit 41).

We also analyzed NHATS to develop prediction equations of how much unpaid care is provided. An example of unpaid care is an informal care giver such as a family member or friend. The purpose of this analysis is to determine if trends affecting future supply and demand for unpaid care might affect future demand for paid care. Demand for unpaid care may affect future demand for direct care workers. The regression estimates from NHATS were applied to the non-nursing home population age 65 and older in the population database to model total hours of unpaid care.

First, using logistic regression, we analyzed the propensity of individuals to use paid and unpaid care (Exhibit 42). Older age and presence of activities of daily living limitations were associated with greater odds of receiving both paid and unpaid care.

Second, we analyzed total paid and unpaid care hours received per week for those individuals who reported receiving at least one hour of care (Exhibit 43). This was done using a negative binomial regression model. FTE demand for unpaid care assumed 1 FTE equal to 40 hours of unpaid care. Older age and presence of select activities of daily living limitations were associated with greater number of paid and unpaid hours per week of care received.

The focus of this work is modeling growth in demand for unpaid hours of care. Projections of growth in paid hours of care were consistent with projected growth in demand for personal care aides and home health aides. We explored whether the trend in decreasing family size might affect future supply of unpaid care and the implications for demand for paid care.4  Analysis of NHATS found that the smaller family size is correlated with greater weekly hours of paid care and fewer weekly hours of unpaid care. However, the overall impact of decreasing family size is relatively small. It also does not appear to substantially affect demand for paid care workers.

The Status Quo scenario for modeling demand assumes prevalence rates of functional impairments among people of different age, sex and race/ethnicity will remain constant over time. It assumes that recent patterns of care use and delivery will remain unchanged. It does, however, incorporate population growth and aging. Demand projections were developed at the state level and then aggregated to obtain the national projections. State-level projections take into consideration geographic variation in health risk factors and demographics.

Demand is modeled under a scenario focusing on forecasting population health status. It also captures trends and expectations in care use and delivery. This Population Health scenario is described in more detail in The Nursing Model Components. It assumes that the nation achieves sustained reductions in excess body weight; smoking cessation; and improving uncontrolled hypertension, hypercholesterolemia, and hemoglobin A1C levels. Such a scenario might be achieved under a medical home model. It is also based on national priorities to improve access to preventive care. Trends that might help achieve such a scenario include:

  • increased organizational and policy commitment to population health as illustrated by
    • health care reform
    • ACO-related quality metrics targeted at population health
    • payment reform
  • greater assumption of risk by providers
  • better infrastructure to manage population health.
Date Last Reviewed: