Abstract

Direct support professionals (DSPs) provide a broad range of supports in a variety of settings to people with intellectual and developmental disabilities (IDD) that enables people to live, work, and participate in their communities. Despite the crucial importance in ensuring supports for community participation of people with IDD, high rates of annual turnover among DSPs in organizations that employ them have been documented for decades. This study utilizes National Core Indicators Staff Stability data from 2016 to examine the impact of organizational- and state-level factors related to DSP turnover, including annual DSP turnover and the percentage of DSPs who left their positions after less than 6 months. At the organizational level, a higher turnover rate in the last 12 months was significantly related to lower DSP wages and to not offering health insurance. At the state level, a higher turnover rate in the last 12 months was significantly related to a lower percentage of people living in individualized settings and lower per capita Medicaid spending. For early turnover at the organizational level, a higher percent of leavers within 6 months of tenure was significantly related to not offering paid time off and health insurance, higher vacancy rates, higher proportion of part-time DSPs, and lower overall staff sizes.

Direct support professionals (DSPs) provide support to people with intellectual and developmental disabilities (IDD) that enables them to live, work, and participate in their communities. DSPs work in a variety of home and community-based settings, including family homes, group homes, and vocational and job sites. They provide a broad range of individualized supports so that people with IDD can access similar experiences and opportunities as people without IDD. Core competencies for DSPs reflect a wide array of skills needed, including support for health and hygiene, identifying and providing person-centered services according to peoples' goals, supporting self-determination, and full and active community participation (Centers for Medicare and Medicaid Services, 2014; National Alliance on Direct Support Professionals, 2016). Supports provided by DSPs are a cornerstone to community participation for people with IDD. Despite the crucial importance of these supports, high rates of annual turnover among DSPs in organizations that employ them has been documented for over 30 years (American Network of Community Options and Resources [ANCOR], 2001; Barry Associates, 1999; Bogenshutz et al., 2014; Hewitt et al., 2000; Lakin & Bruininks, 1981; Oklahoma Developmental Disabilities Service Division, 2000; Test et al., 1999). The shortage of a qualified, competent, and stable DSP workforce curtails peoples' opportunities in the community. As such, this long-term, sustained workforce shortage was recently identified as a pervasive systems failure in long-term services and supports (Hewitt et al., 2018). Solutions to address high rates of DSP turnover are critical. The purpose of this secondary data analysis is to examine the impact of organizational- and state-level factors related to turnover of DSPs using data from the National Core Indicators Staff Stability Survey. It describes and evaluates trends in the rate of overall DSP turnover during 2016, and the percentage of those DSPs who left their position in less than 6 months. These findings can be used to inform policy, identify areas in need of further analysis for potential improvement activities at the organizational and state level, and identify opportunities for change that could stabilize this vital workforce.

The direct support workforce is one of the fastest growing occupations in the United States (Bureau of Labor Statistics [BLS], 2015; Paraprofessional Healthcare Institute Policy Works [PHI], 2014). To support community living for people with IDD, it is estimated that an additional 900,000 full-time equivalent DSPs will be needed annually by the year 2020 (Hewitt, 2013), and the problem is expected to become even larger in coming years (National Academies of Sciences, Engineering, and Medicine, 2016). The average annual turnover rate in 20 states and the District of Columbia for organizations that support people with IDD is 45.5% (Human Services Research Institute & The National Association of State Directors of Developmental Disabilities Services [HSRI & NASDDDS], 2018; National Core Indicators [NCI], 2018). This can be compared to the national average of 3.4% total separations across all industries in December 2016 (BLS, 2017). According to National Core Indicators Staff Stability Survey, of those DSPs who left their positions, 38% had been employed in their position for less than 6 months, and an additional 21% had been employed in their position between 6 to 12 months when they left. This means that nearly 60% of the DSPs who left in 2016 had been employed for less than 1 year in their position. These high rates of turnover are costly on a number of levels. Turnover affects real costs, increasing the amount of time agency managers must spend recruiting, interviewing, and onboarding new staff and overtime costs for staff who are working additional time to cover the hours the provider is obligated to offer. It also has emotional, social, and other costs for those who receive supports and for loved ones who often take up support responsibilities when no qualified DSP can be found.

Costs of Turnover at Individual and Organizational Levels

People who rely on supports provided by DSPs report experiencing lower quality of life when stable, reliable supports cannot be found. A study of 1,342 people with IDD who participated in a Council on Quality and Leadership (CQL) Personal Outcomes Measures interview analyzed the impact of DSP turnover in the previous 2 years on 21 indicators that are aligned with quality of life (Friedman, 2018). DSP turnover impacted 17 of 21 outcome areas, including people's sense of safety, health, interactions with others in the community, participation in community life, friendships and intimate relationships, being treated with fairness and respect, developing natural supports, choice in where and with whom to live, and choice in services, among other areas. Fifty-six percent of people had experienced DSP turnover in the last 2 years. An earlier study among people receiving supports found that DSP turnover increased peoples' feelings of sadness, and that they found it difficult to develop trust and trusting relationships with new DSPs consistently entering their life (Test et al., 2003). People reported feelings of responsibility and self-blame when DSPs left their job.

When DSP staff cannot be recruited, hired, and maintained to provide supports for people who require them to live and work in the community, provision of supports can fall on family members and loved ones. Family members have reported taking on support responsibilities, and missing work and other opportunities when qualified and reliable DSPs cannot be found (Anderson et al., 2011; President's Committee for People With Intellectual Disabilities [PCPID], 2017). Inconsistencies in supports can curtail career opportunities and even the ability to maintain employment if direct support for their loved one cannot be found. Using the National Health Interview Survey Disability Supplement, Anderson and colleagues (2002) found that 53% of parents reported not taking a job, working fewer hours, leaving the workforce, turning down a promotion, and other major career concessions as a result of caring for a child with IDD. Furthermore, supports are increasingly provided in family homes (Larson et al., 2018), which can leave families unable to leave home to attend work or any other community opportunities when reliable DSPs cannot be found to support their loved one.

A high rate of DSP turnover is also costly for human service organizations, and state budgets more broadly. In 2004, Larson and colleagues (2016) estimated turnover costs at $3,278 per DSP who leaves the organization, including separation costs, training costs, and hiring costs. In 2009, these costs were estimated at $4,872 per DSP (ANCOR, 2010). These aggregated costs are substantial, with turnover rates nearing 50% of the workforce annually and costs related to turnover estimated at over $2.3 billion nationally (PCPID, 2017). Efforts to understand factors related to turnover on both organizational and state levels can contribute to problem-solving efforts to stabilize the DSP workforce.

National Core Indicators Staff Stability Survey

Starting in 2014, National Core Indicators (NCI), an initiative of the National Association of State Directors of Developmental Disabilities Services (NASDDDS) and the Human Services Research Institute (HSRI), began collecting organizational-level data on the direct support workforce (National Core Indicators [NCI], 2014). The NCI Staff Stability Survey was developed in collaboration with the National Direct Service Resource Center to assess critical and relevant information about DSP workforce stability, wages, benefits, and recruitment and retention strategies (NCI, 2014). The staff stability dataset includes DSPs employed by agencies of all types (private for-profit, private nonprofit, county/local government run, state-operated, etc.)

This study is the first to examine trends in organizational- and state-level factors related to DSP turnover using data from the NCI Staff Stability Survey. The purpose of this study was to investigate organizational factors, including: wages, number of annual overtime hours per DSP, provision of paid time off, offerings of health insurance, and the ratio of part-time DSPs employed in the organization on two outcomes. The outcomes were annual DSP turnover and early separations, defined as the percent of DSPs who left the organization with experience equal to 6 months or less. State-level factors, including state unemployment rate; percentage of people with IDD living in individualized, family, and institutional settings; per capita Medicaid spending; and participation in Medicaid expansion, were also analyzed.

Organizational-Level Predictors

DSP Wages

The average hourly DSP wage in the National Core Indicators 2016 Staff Stability Survey was $11.76 per hour (HSRI & NASDDDS, 2018). For full-time work, according to the U.S. Department of Health and Human Services (2019), this was below the federal poverty line for a family of four that same year. Several studies have found that over half of all DSPs utilize public benefits because they cannot meet their basic expenses (Hewitt et al., 2019; Paraprofessional Healthcare Institute [PHI], 2015). Furthermore, wages paid to the direct support workforce have not increased in over a decade when adjusting for inflation (Campbell, 2018). Despite years of experience in the field, DSPs wages paid have not kept up with the increased level of skill developed through the years of service compared to new employees hired in the organization. The average hourly starting pay for DSPs in 2016 was $10.79 (HSRI & NASDDDS, 2018). In a secondary data analysis of factors related to turnover of 2,221 certified nursing assistants in nursing facilities, an additional dollar per hour in wages was associated with 2.1 added months of tenure (Wiener et al., 2009). Powers and Powers (2010) found that a 10% increase in pay for DSPs providing supports for people with IDD reduced organizational turnover rates by three percentage points, a similar result to the ANCOR wage study that found that every $1 increase in DSP starting wage predicted a 3.61% decrease in DSP turnover (Anderson-Hoyt et al., 2010). With turnover at 45.5%, on average, assuming increasing wages by 10% leaves a continuing turnover rate over 40%. Although wage increases are an important step, further analysis of additional factors contributing to high turnover could assist with overall improvement of this crucial issue.

Number of Overtime Hours

The DSP workforce has experienced high vacancy rates (9.8% among full-time DSP positions and 15.4% among part-time DSP positions in 2016; HSRI & NASDDDS, 2018). This can mean that other DSPs cover open shifts in order to provide needed supports for people. The average number of overtime hours paid to direct support workers in Minnesota in 2017 was 8,047, with annual overtime hours per DSP at 38 (Pettingell et al., 2019). Sometimes organizations expect supervisors to work extra hours to fill shifts that are open due to unfilled positions (Pettingell et al., 2019), but other times DSPs can work these hours and earn overtime pay at a rate of time and a half. Research on the impact of overtime hours on turnover is scarce. Overtime hours may be an important source of income that keeps dedicated DSPs in their job by earning a higher rate of pay. Test and colleagues (2003) found that over one-third of DSPs work a second or third job. A recent study in Minnesota of the direct support workforce found that over 28% work at least one additional job, with some working three additional jobs (Hewitt et al, 2019). However, it may also lead to supervisor and/or DSP exhaustion by working extra hours. Further, some settings where people receive supports may compel organizations to pay a higher number of overtime hours. Group homes and institutional settings may require overtime support, creating an urgency among management to fill vacant shifts to ensure staff are available; home settings where people receiving supports live with their family or in their individual homes may not create the same level of urgency among managers because family steps in when DSPs are not available (Anderson et al., 2011).

Paid Time Off

Organizations provide access to paid time off for DSPs in a variety of ways, including paid sick time, paid vacation time, and general availability of paid time off that is not specifically tracked to illness or vacation. In 2016, 35.2% of agencies offered general paid time off (not tracked to type) to all DSPs, and an additional 33.3% offered general paid time off to full-time DSPs. In addition, another 44.7% of agencies offered paid sick time off to either full-time or all employees, and 49.8% reported offering paid vacation time off to either full-time or all employees (HSRI & NASDDDS, 2018). Among direct support workers, there are indicators of workplace burnout (Skirrow et al., 2007), especially among full time workers (Vassos et al., 2012). Provision of paid time off may be an important way of keeping DSPs in their job when they experience illness, exhaustion, burnout, or other personal situations.

Access to Health Insurance Benefits

Similar to providing paid time off, health insurance is another benefit that providers report difficulty making accessible to DSPs based on costs. In 2016, only 14.5% of providers offered health insurance to all employees and 66% offered health insurance benefits to full-time DSPs only (HSRI & NASDDS, 2018). In total, this means 80.5% of the 2,310 agencies reported offering health benefits to some or all employees. This is down from a study in 16 sites providing residential and day services to people with IDD in 2009–2011 in Minnesota, where all organizations reported offering health care benefits to some employees, although the number of hours employees had to work to obtain the benefits varied considerably across site (Bogenshutz et al., 2014). Yet, a recent study demonstrated that, even when health insurance is offered to employees, only about 38% actually utilize the benefits (Pettingell et al., 2019). Benefits may be too costly for DSPs to afford; as an alternative, they may access benefits through public assistance or another source, such as coverage offered through a spouse's employer or through a second job. This warrants deeper understanding of the experience among DSPs across the country. It is important to note that benefits, along with wages, can vary widely from state to state and among providers within each state.

Percentage of Part-Time DSPs

The direct support workforce is comprised of a large percentage of part-time workers. About 1/3 of the direct support workforce employed by organizations that participated in the Staff Stability Survey were part-time employees (HSRI & NASDDS, 2018). In Minnesota in 2017, over half of direct support workers were part-time positions (Pettingell et al., 2019). In both of these studies, rates of turnover are higher among part-time DSPs, and part-time status may limit availability of benefits and make it more difficult to consider direct support as a career or long-term employment option.

Covariates

Vacancy Rates

There is an increased demand for community-based supports and, as a result, this field is struggling to recruit enough workers to keep up with the demand. Demographic and economic changes in the United States, including an aging population and competitive wages, are contributing factors to high vacancy rates for DSP positions in long-term services and supports. High vacancy rates may mean that organizations are unable to meet the demand for services (Hewitt et al., 2015), and people receiving supports are unable to live fully in the community, including going to work, participating in community activities and enjoying time with friends and family. Vacancy rates among full-time DSPs was 9.8% in 2016, but part-time vacancy rates were higher at 15.4% during the same period (HSRI & NASDDDS, 2018).

Overall Staff Size

The size of an organization and number of staff may mean that there are different demands placed on DSPs. The number of employees on organizations' payrolls vary greatly within and across states, with a range of 35.6–214.0 staff, average staff size of 102.5, and median of 54.3 among the participating states in 2016 (HSRI & NASDDDS, 2018). Powers and Powers (2010) found that an increase of 20 employees on staff in an organization increased turnover rates between 29–40%. The size of an organization may also impact benefit options available for the staff.

Number of Types of Services Provided

Organizations that deliver home and community-based services provide differing types of services, including residential, in-home supports, and day or employment services. Some organizations provide only one type of service, others provide two or three types. Wages in nonresidential and vocational programs supporting people with IDD have been historically higher (e.g., Bogenshutz et al., 2014), and ratios of part-time workers differ across these service types. In 2016, DSPs providing nonresidential supports in day or work programs earned an average of $12.04, DSPs providing in-home supports earned $11.62, and DSPs providing residential supports earned $11.41 (HSRI & NASDDDS, 2018).

State-Level Predictors

State-level differences in policies and practices can have an impact on the delivery of services (Larson et al., 2018). Houseworth, Stancliffe, and Tichá (2018) utilized multilevel modeling to demonstrate the impact of state-level differences in addition to individual-level differences on aspects of people's choices when examining 2013–14 National Core Indicators data from the Adult and Consumer Survey. State-level factors were strongly related to support-related choices, including with whom and where the person lived, staff members and case managers supporting the person, and the types of day or work activities in which the person participated. For example, a higher proportion of people with IDD living independently in a state predicted more support-related choices. Several state-level factors are examined in this article to quantify the importance of state policy differences and to detect trends in DSP turnover related to state-level policies, specifically state unemployment rate; percentage of people with IDD living in individualized, family, and institutional settings; per capita Medicaid spending; and participation in Medicaid expansion.

State Unemployment Rate

The economic recession that began in 2007 resulted in the loss of employment opportunities and temporarily created a larger pool of applicants for DSP jobs. As recovery occurred, the pool of applicants for DSP positions has decreased, as has the quality of applicants, given that applicants have many options for employment. Among human service workers, a higher unemployment rate has been positively associated with longer tenure in this workforce (Powers & Powers, 2010; Wiener et al., 2009). Rising vacancy rates in the last several years certainly could be associated with lower state unemployment rates.

Settings Where People With IDD Live (Individualized, Family, and Institutional Settings)

More people in the United States in recent decades have received services and supports outside of institutional settings than in previous generations (Larson et al, 2018). Across states, of those people who received long-term services and supports in 2016, 58% lived in the home of a family member, 25% lived in a group home or institutional care facility for people with IDD, 12% lived in a home they owned or leased, and 5% lived in a host or foster home (Larson et al., 2018).

Per Capita Medicaid Spending

Differences between state spending can be attributed to population density differences between states. One way to examine these differences is to divide the total amount of a state's Medicaid spending on the state's overall population to get the per capita Medicaid spending by state. States vary greatly in these expenditures, from $292 to $17, with an average of $110 (Larson et al., 2018). This may be an influencing factor on turnover and vacancy rates of the DSP workforce. Note we defined this in terms of state population, not the service population.

Participation in Medicaid Expansion

The Supreme Court's 2012 ruling (National Federation of Independent Business v. Sebelius) on the Affordable Care Act (ACA) allowed states to opt out of the law's Medicaid expansion, which was designed to change the financial eligibility requirements used to determine Medicaid eligibility to allow states to cover previously ineligible groups, namely the working poor. The ruling left the decision to expand or not to each state, largely decided by governors and/or state legislators. As of 2019, 33 states and the District of Columbia had expanded Medicaid. In 2018, 30% of direct support workers in Minnesota indicated that they rely on a government-paid program (Medicare, Medicaid, Tricare) for health insurance, and 71% utilize some sort of public or government-paid health benefits (Hewitt et al., 2019). Using American Consumer Survey data from 2016, 39% of U.S. home care workers accessed health insurance through Medicaid, Medicare, or another public source, and 18% lacked health insurance altogether (Campbell, 2018).

Purpose of the Current Study

The purpose of the current study is to examine the impact of organizational- and state-level factors on DSP turnover by conducting secondary data analysis of a random sample of 1,406 organizations that provide supports and services to adults with IDD in 20 states and the District of Columbia using the NCI 2016 Staff Stability Survey (HSRI & NASDDS, 2018). This study seeks to describe and evaluate trends in: (1) the rate of overall DSP turnover in 2016; and (2) the percentage of DSP early separations, defined as those who left their position in less than 6 months from their date of hire. Organizational- and state-level factors related to each are examined. The two research questions included:

  1. What are the organizational-level predictors of turnover, as controlled for by vacancy rate, number of services provided, and organizational size?

  2. What state factors are related to organizational-level predictors of turnover, as controlled for by vacancy rate, number of services provided, and organizational size?

Method

Instrument: NCI 2016 Staff Stability Survey

National Core Indicators (NCI) is a combined effort of 20 years between the National Association of State Directors of Developmental Disabilities Services (NASDDDS) and the Human Services Research Institute (HSRI). Their work measuring and tracking results by state for adults ages 18 and older with IDD began in 1997. In 2013, they began an additional effort with a small group of states to collect data on DSP vacancy and turnover rates among provider agencies that deliver direct support to adults with IDD through their NCI Staff Turnover survey tool. That tool was converted to a more comprehensive survey focusing on DSP workforce issues (e.g., benefits, wages, retention, and recruitment activities). Input was gathered from experts in the field, two rounds of focus groups were conducted with provider agencies, and a pilot test was conducted in two states in 2014 for the purposes of refining the survey. It was first released to collect 2014 data (NCI, 2014). Data are used to provide information to states that will help reduce the costs related to staff turnover, as well as the impact turnover has on the quality of the supports consumers receive. An increasing number of states participate in the NCI Staff Stability Survey since its pilot in 2014 (NCI, 2014). The data also provide a mechanism to measure state improvement. The 2016 Staff Stability Survey (HSRI & NADDDS, 2018) asks questions of the leadership of provider organizations in participating states related to DSP current staff, staff separations, status (full- or part-time) of staff, hours worked, wages, and benefits.

Sample

State Selection

Data were collected from 20 states and the District of Columbia that voluntarily participated in the NCI program in 2016. States included: AL, AZ, CT, GA, HI, IL, IN, MD, MO, NE, NY, OH, OR, PA, SC, SD, TN, TX, UT, VT, and DC. Ohio collected data separately for intermediate care facilities for people with intellectual disability (ICF/ID)-funded supports and Home and Community Based Services (HCBS) waiver-funded supports. These data were combined for the current analyses. Data were gathered for the time period from January 1, 2016, to December 31, 2016. Sampling methods varied for each state; however, most states surveyed all agencies providing direct support services to adults with IDD. See Appendix B in the 2016 Staff Stability Report for specific state methods (HSRI & NADDDS, 2018).

Participants

Across the states, 3,222 provider agencies completed the 2016 Staff Stability Survey. Due to missing data, the listwise number of providers assessed for this study was 1,410 for early turnover and 1,618 for annual turnover from 20 states and the District of Columbia.

Variables

Table 1 shows descriptive statistics on all variables in this study. The outcomes examined were annual DSP turnover, and percentage of those DSPs who were employed in their position for less than 6 months. Turnover was computed using a formula for annual crude separation rate, which is calculated using three variables: the number of DSPs who left their position for any reason in the calendar year divided by the total number of DSPs employed plus the number of DSP vacancies. This is multiplied by 100 to yield a percentage of DSP turnover. This formula has been used widely in studies within and outside of the long-term services and supports industry (Bogenschutz et al, 2015, Larson & Sedlezsky, 2015). This computation differs from turnover reported in NCI Staff Stability analyses, as vacancy rate is not included in their denominator. As seen in Table 1, at the organization level, average DSP turnover in the last 12 months was 24.58%.

Table 1

Description of the Sample Organizations (N = 1,618)

Description of the Sample Organizations (N = 1,618)
Description of the Sample Organizations (N = 1,618)

Early separations turnover in this study is defined as the percentage of DSPs who turned over in the previous year and who had been employed in their position for less than 6 months from their date of hire. Early separations was computed by dividing the number of DSPs who left their position within 6 months of hire in the calendar year by the total number of DSPs who left their position that same year. This number was multiplied by 100 to yield a percent. Of those DSPs who left their positions, an average of 19.58% left within 6 months of hire.

Five organizational level predictors of the outcomes were examined. Descriptive statistics for these are also presented in Table 1.

Wages

This was a single item asking the average wages for DSPs across all services and settings. The average DSP hourly wage was $11.68.

Percent Providing Paid Time Off

Individual items asked if the organization provided eligible DSPs paid time off, paid sick time, and/or paid vacation. If an answer of “yes” was provided for any of these items, a code of 1—“yes, offers for paid time off”—was assigned. If a yes answer was not provided, a code of 0—“no, does not offer paid time off”—was assigned. Eighty-six percent organizations offered paid time off.

Overtime Hours per DSP

This was a single item asking the total number of overtime hours paid to DSPs between January 1, 2016, and December 31, 2016, which was divided by the total number of full- and part-time DSPs employed over that time period. The average number of overtime hours was 138.97 hours per DSP in the calendar year. The Staff Stability survey does not distinguish between hours over 40 that full-time employees may work or hours beyond a typical week part-time employees may work. Therefore, the numbers reported by providers may reflect either or both.

Percent Offering Health Insurance

This was a single item asking if the organization offered eligible DSPs health insurance coverage. If an answer of “yes” was provided, a code of 1—“yes, offers health insurance coverage”—was assigned. If a yes answer was not provided, a code of 0—“no, does not offer health insurance coverage”—was assigned. Seventy-six percent of organizations offered health insurance to some or all DSPs.

Percent of Part-Time DSPs

This was created dividing the number of part-time DSPs by the total number of DSPs. On average, 35% of DSPs in an organization worked part-time.

Three covariates were examined.

Vacancy Rate

DSP vacancy rate was calculated using two variables: number of DSP vacancies and total number of DSPs employed. Vacancy rate was computed by dividing the total number of full- and part-time funded positions currently vacant by the total number of funded positions at the site (the number of DSPs employed plus the number of vacant positions). The result was multiplied by 100 to yield a percent (Larson & Sedlezsky, 2015). The average vacancy rate was 10%.

Number of Service Types Provided

Each organization was asked if it provided: (1) residential, (2) day/employment, and/or (3) nonresidential supports and services. This variable was a count ranging from 0 to 3 types of services provided. The average number of service types provided was 1.49.

Staff Size

This was a single variable that asked the total number of staff on the organization's payroll as of December 31, 2016. The average staff size was 213.73. This includes different kinds of staff, such as administration, human resource, clinical, nursing, and others that were on the organization's payroll on this date.

Six state-level predictors were considered for these analyses.

Residence Size

Three variables were created regarding the number of people and/or setting, where a person who receives services lives. These variables were drawn from the Residential Information Systems Project (RISP) data (Larson et al., 2018). Three variables were: the percent of individuals living in intermediate care facilities (ICF) organizations with 16 or more individuals with IDD and those living in group homes with more than four other individuals, the percent of individuals with IDD who live in family homes, and the percent of individuals with IDD who live in individual or small group settings (e.g., own home, agency apartment, etc.; 1–3 people). At the state level, the average number of individuals living in ICFs or large group settings, individualized, and family settings was 17.86%, 31.71%, and 48.52%, respectively.

Per Capita Medicaid Spending

This variable came from RISP data (Larson et al., 2018). It represents annual Medicaid expenditures per state resident (including home and community-based services and institutional care facility services). The average annual per capita spending by state was $163.90.

Unemployment Rate

This variable comes from the American Community Survey (ACS), a data collection project completed yearly by the U.S. Census Bureau (2016). One-year estimates of unemployed individuals of working age living in the state during 2016 from ACS was used to determine this variable. The average unemployment rate was 6.83%.

Medicaid Expansion

Data from the Advisory Board, a global healthcare think tank in operation since 1979 (https://www.advisory.com/daily-briefing/resources/primers/medicaidmap) were used to determine this variable. If a state included in our analysis expanded Medicaid before January 2016, a code of “1” was assigned. If a state opted out of Medicaid expansion or did not expand before January 2016, a code of “0” was assigned. Just over half (57%) of states had accepted Medicaid expansion by 2016.

State Weight

There were five states (AL, AZ, CT, PA, and TX) that did not meet the margin of error criteria established by the Human Services Research Institute (HSRI) that administers the NCI. HSRI used a sampling estimation program (http://www.raosoft.com/samplesize.html) based on the number of providers within each state to create an estimate of the minimum number of provider surveys completed in order to reach a 95% confidence level. In order to include as many states as possible, a weighting variable was created, i.e. the percent of the surveys completed that met this minimum required number. For example, the lowest response percent was 16%, which indicates that only 16% of the required surveys were completed. By weighting the model using this percentage, this study was able to temper the results from states with low response rate (thus mitigating their impact on overall estimates), but still include their data in the study analysis. The average state weight in this study was 0.78 (SD = 0.20, range 0.16–1.00).

Analysis

Descriptive statistics and hierarchical linear modeling (HLM) were computed using HLM 7.0 (Raudenbush et al., 2013). HLM models were used to examine the effect of predictors across two levels of analysis on the two workforce outcomes of early and annual turnover. Restricted maximum likelihood (RML) was used as the estimation method because of the relatively small sample size for the level-two predictors (Raudenbush & Bryk 2002). The final analysis was based on 20 states and the District of Columbia. All analyses were evaluated using α = 0.05. All models used robust standard errors.

Multilevel models incorporate higher-level predictors by taking into account the contribution of each level of analysis (Raudenbush & Bryk, 2002). This process allows for the explanation of variability due to group or cluster effects by modeling lower-level intercepts and slopes based on group-level variables. HLM accounts for the interdependence of observations by partitioning the variance of a dependent variable, such as turnover rates, into hierarchical levels. Therefore, it improves estimation of parameters given by the GLM approach (Raudenbush & Bryk, 2002). This improvement is due to variance being explained by characteristics of providers (level 1) and variables specific to states (level 2).

Results

This study attempted to fit state-level variables at level 2 of HLM for both annual DSP turnover and early turnover. For early turnover, the model did not converge and the exhausted attempt at estimating level 2 (state level) variables produced a model fit that was below the intercept-only model and with an overall model significance level of about .50. These results strongly suggest that level 2 estimates are likely random for this model. Thus, only level 1 estimates for early DSP turnover are presented.

As seen in Table 2, at the organizational level, a higher percent of leavers within 6 months of tenure is significantly related to less offering of paid time off (PTO; p = 0.025), less offering of health insurance (p < 0.001), higher vacancy rates (p < 0.005), higher proportion of part-time DSPs (p < 0.001), and lower overall staff sizes (p = 0.014). PTO and health insurance represent the organization offering the option, not whether the employees were eligible or had taken up the offer.

Table 2

Hierarchical Linear Model on Early DSP Turnover (Intercept Only)

Hierarchical Linear Model on Early DSP Turnover (Intercept Only)
Hierarchical Linear Model on Early DSP Turnover (Intercept Only)

For annual turnover, the level-2 model converged and produced acceptable model fit. Thus, the full model for turnover is reported.

As seen in Table 3, at the organizational level, a higher turnover rate in the last 12 months is significantly related to lower DSP wages (p < 0.001) and not offering health insurance (p = 0.037). At the state level, a higher turnover rate in the last 12 months is significantly related to a lower percentage of people living in individualized settings (p = 0.049) and lower per capita spending (p < 0.001).

Table 3

Hierarchical Linear Model for Annual DSP Turnover

Hierarchical Linear Model for Annual DSP Turnover
Hierarchical Linear Model for Annual DSP Turnover

Annual DSP turnover and early turnover were both predicted by whether or not an organization offered health insurance for DSPs.

Discussion

This study revealed several important trends related to annual and early turnover of DSPs employed in their position for 6 months or less. Specifically, there were different factors related to DSP turnover in each model based on DSP tenure. Not surprisingly, higher wages and the option of accessing health insurance were related to lower annual turnover rates. A higher percentage of part-time DSP positions in an organization and the organization's offering paid time off were also related to higher early turnover rates. Although these are initial results based on 1 year of data, they can inform rate setting in long-term services and supports as well as hiring and compensation practices in human service organizations. Efforts to better retain DSPs will necessitate increased state and federal spending to increase wages, and access to full-time work hours, health insurance, and PTO benefits.

In the early separation DSP model, a lack of provision of paid time off and health insurance and having a higher percentage of part-time DSP positions in the organization were related to a higher percentage of DSPs who turned over in 6 months or less. Organizations who had a higher percentage of part-time workers also had a higher early turnover rate, although this was not the case for annual turnover. Part-time DSPs may be more likely to leave their position if they are able to find full-time employment elsewhere, using direct support as a bridge to finding other employment. Organizations have reported that they have a more difficult time retaining DSPs as the unemployment rate decreases (PCPID, 2017; Powers & Powers, 2010). Developing strategies and service models that maximize the use of full-time workers and pooled health insurance options may be important strategies for retaining DSPs beyond 6 months.

These findings also align with the need to provide health insurance benefits and PTO to part-time workers. States included in the 2016 NCI Staff Stability Survey report that around one-third of organizations that provided pooled paid time off or paid sick or vacation time to DSPs offer this to full-time DSPs only (HSRI & NADDDS, 2018). In a recent study of direct support workers in Minnesota, finding a job with access to higher wages or benefits were top reasons that workers reported that DSPs left their positions (Hewitt et al., 2019). Relative to the demands of direct support, other industries that pay similar wages but offer more attractive benefits packages may sway workers from direct support into other industries.

Provision of health insurance was a factor related to higher rates of DSP turnover in both the early turnover model and the annual turnover model. New federal regulations requiring people to have health insurance went into effect in 2010 (Patient Protection and Affordable Care Act) and people who did not carry health insurance faced a penalty on their taxes, if they were not able not verify that they had health insurance. This law may have played a role in these findings. Furthermore, many DSPs turn to government programs for health insurance. In a recent study in Minnesota, 30% of workers reported they obtained health insurance from government programs. PHI Policy Works (2015) reports that 52% of workers utilize government health benefits. It is important to pursue a longitudinal analysis to be able to better gauge the impact of this new policy on DSP turnover rates.

It is unsurprising that higher wages are related to a lower percentage of overall DSP turnover. Increasing wages has been shown to be related to lower DSP turnover in previous studies (ANCOR, 2010). Wage rates paid to DSPs have been tracked in recent decades to show that rates of pay have not kept up when adjusting for inflation (Espinoza, 2017). Livable wages take into account geographically specific costs of living to sustain a family's basic needs, including housing, food, childcare, transportation, health insurance, and other necessities, without reliance on government subsidies. For example, the #bFair2DirectCare campaign has advocated for livable wages paid to DSPs in New York state. This campaign has utilized the Massachusetts Institute of Technology (MIT) living wage calculator to determine differences in wages across geographic regions in the state (Nadeau & Glasmeier, 2018). Advocacy for higher wages cannot merely be focused on raising wages, but, instead, must focus on increased wages at a competitive rate to sustain this workforce over time. Without livable wages, DSPs may want to stay in this line of work, but find the rate of pay unsustainable to cover living costs over time. This may result in DSPs finding a different line of work with higher pay when they would have otherwise stayed in their job. Organizations can also support established DSPs by creating career pathways that reward increased level of skills with commensurate increases in pay (e.g., Espinoza, 2019; Hewitt et al., 2015). Given the high costs of recruiting and training new direct support staff (ANCOR, 2010; Larson et al., 2016), investment in training and recognizing committed DSPs are important considerations for organizations.

In the United States, there are numerous service industries. These include sanitation, hospitality, environmental protection, and many others. The wages paid to these types of service workers are often higher than those of DSPs. For example, average hourly wages in 2018 for restaurant cooks were $13.26, waiters and waitresses were $12.42, janitors and cleaners were $13.92, landscaping and groundskeepers were $14.88, and retail sales workers were $13.61 (Bureau of Labor Statistics, 2019). When unemployment is low and other service jobs are paid higher rates, the ability to recruit people into DSP jobs becomes more challenging. Direct support can be challenging and take a higher level of skill than such jobs that offer higher wages. These competition factors need to be considered when determining rates for long-term services and supports (LTSS) in the intellectual and developmental disability field.

State Factors

Regarding the second research question, what state factors related to organizational-level predictors of turnover, as controlled for by vacancy rate, number of services provided, and organizational size, these factors were only modeled on overall DSP turnover. Both the higher percentages of individuals living in individualized settings and higher per capita spending within a state were associated with less turnover.

A greater number of people living in individualized settings were related to lower overall turnover rates. Individualized settings are those with three or fewer people living in them. Houseworth, Stancliffe, and Tichá (2018) found that people living in their own home or apartment was associated with more choice in support options for individuals with IDD. Increased choice is desirable for people receiving supports, and in line with policies (e.g., HCBS Settings Rule). It also seems that there is less turnover when DSPs work in more individualized settings, which may be related to having the opportunity to really get to know the individuals, being able to plan and implement more individual person-centered supports and having less stress with fewer people for whom the DSP has a high level of responsibility and risk. State systems have been moving continuously toward more individualized supports, and continuing this movement may be important for states to address issues of DSP turnover as well.

Higher per capita spending was also associated with less turnover. The states with lower turnover may also be paying higher wages to DSPs. It may also be that per capita spending leads to more consumer satisfaction, which makes working as a DSP more rewarding. It is clear that state (and subsequent federal) investment in long-term services and supports will be needed if wages for DSPs are able to increase.

Contrary to previous studies, these analyses did not find that state unemployment rate was related to turnover (Powers & Powers, 2010; Wiener et al., 2009). It seems a logical assumption that, when unemployment rates are low, there are fewer workers available to work across all industries. Yet, other factors influence satisfaction and commitment to stay at a job. The demands and accountability aligned with training and credentials for the job, access to and effective support from a supervisor, conflicts with co-workers, and changing demographics are all contributing factors, as are wages and benefits (Bogenshutz et al., 2015; Hewitt et al., 2019; Lerman et al., 2014; Wiener et al., 2009). Perhaps these other factors that are known to influence retention are influencing factors here.

Conclusion and Future Direction

This is the first study using NCI Staff Stability Survey data for systematic analysis across multiple states. The results likely indicate that wages and benefits (PTO and health insurance) are critical factors to retaining DSPs in their positions. These are important factors for both states and organizations to address. Yet, further analyses are needed over a longitudinal period to better understand these influencing factors. Additional variables likely to influence retention and turnover must also be included in future studies. Further analyses should examine the uptake of health insurance by DSPs, and survey where DSPs are receiving health insurance, if not from their employer. This analysis should consider how the average monthly premium for individuals and family coverage may be preventing health insurance uptake. Another area of examination includes answering why there is lower turnover of staff working in more individualized settings, yet there was no significant difference for staff in ICFs and working in family settings. Future studies should consider analyzing the relationship between average spending per service user, as this may be more related to the workforce. Similarly, it would be important to consider whether wages within a state systematically increase with higher spending. Finally, an analysis with a larger number of states will provide information that is more generalizable and enable for more robust multilevel analyses. Also, we defined per capita spending in terms of the state population.

Limitations

The small sample size within many states and the low number of states included in the study lead to low fit statistics for 2-level models. More states and providers could make other variables become significant, especially at the state level. It is unclear as to the degree to which providers are actually providing health insurance for their DSP staff. The NCI Staff Stability Survey merely ask if it is offered, not the number or percent of DSPs enrolled who actually use the benefit. The study employed a correlational cross-sectional analysis. Therefore, causal claims cannot be made. The degree to which changes in these factors over time affect staff stability are unknown.

References

References
American Network of Community Options and Resources
.
(2001)
.
State of the states report
.
Author
.
American Network of Community Options and Resources
.
(2010)
.
State of the states report
.
Author
.
Anderson,
L,
Larson,
S. A.,
Lakin,
K. C.,
&
Kwak,
N.
(2002)
.
Children with disabilities and social roles: An analysis of the 94/95 NHIS-D
.
DD Data Brief
,
4
(1)
.
Anderson,
L. L.,
Larson,
S. A.,
&
Wuorio,
A.
(2011)
.
2010 FINDS National Survey technical report part 1: Family caregiver survey
.
University of Minnesota, Research and Training Center on Community Living
.
Anderson-Hoyt,
J.,
McGee-Trenhaile,
M.,
&
Gortmaker,
V.
(2010)
.
Direct Support Professional Wage Study: 2009
.
ANCOR
.
Assistant Secretary for Planning and Evaluation
.
(2019)
Computations for the 2016 poverty guidelines 04/25/2016
.
Barry Associates
.
(1999)
.
The Ohio provider resource association 1999 salary and benefits survey
.
Ohio Provider Resource Association
.
Bogenschutz,
M. D.,
Hewitt,
A.,
Nord,
D.,
Hepperlen,
R.
(2014)
.
Direct support workforce supporting individuals with IDD: Current wages, benefits, and stability
.
Intellectual and Developmental Disabilities
,
52
(5)
,
317
29
.
Bogenschutz,
M.,
Nord,
D.,
&
Hewitt,
A.
(2015)
.
Competency-based training and worker turnover in community supports for people with IDD: Results from a group randomized controlled study
.
Intellectual and Developmental Disabilities
,
53
(3)
,
182
195
.
Bureau of Labor Statistics
.
(2015)
.
Economic news release: Fastest growing occupations 2014-24
.
Bureau of Labor Statistics
.
(2017
,
February)
.
Job openings_and_labor turnover–February 2017 (News Release USDL-17-0179)
.
Bureau of Labor Statistics
.
(2019)
.
May 2018 national occupational employment and wage estimates United States
.
Author
.
Campbell,
S.
(2018)
.
U.S. homecare workers: Key facts
.
PHI
.
Centers for Medicare & Medicaid Services
.
(2014)
.
CMS direct service workforce core competencies final competency set
.
National Direct Service Workforce Resource Center
.
Espinoza,
R.
(2017)
.
U.S. home care workers: Key facts
.
Paraprofessional Healthcare Institute
.
Espinoza,
R.
(2019)
.
Workforce matters: The direct care workforce and state-based LTSS social insurance programs
.
Paraprofessional Healthcare Institute
.
Friedman,
C.
(2018)
.
Direct support professionals and quality of life of people with intellectual and developmental disabilities
.
Intellectual and Developmental Disabilities
,
56
(4)
,
234
250
.
Hewitt,
A.,
Larson,
S. A.,
&
Lakin,
K. C.
(2000)
.
An independent evaluation of the quality of services and system performance of Minnesota's Medicaid Home and Community Based Services for persons with mental retardation and related conditions
.
University of Minnesota, Research and Training Center on Community Living
.
Hewitt,
A.,
Macbeth,
J.,
Merrill,
B.,
&
Kleist,
B.
(2018)
.
The direct support workforce crisis: A systemic failure
.
Impact, 31
.
Hewitt,
A.,
Taylor,
M.,
Kramme,
J.,
Pettingell,
S.,
&
Sedlezky,
L.
(2015)
.
Implementing direct support professional credentialing in New York: Final technical report
.
Office for People with Developmental Disabilities
.
Hewitt,
A.,
Pettingell,
S.,
&
Kramme,
J.
(2019)
.
Minnesota Direct Support Worker Survey 2019: Final report
.
Institute on Community Integration, Research and Training Center on Community Living
.
Hewitt,
A.,
Stancliffe,
R.,
&
Emerson,
E.
(2013)
.
Community living and participation: Embracing the complexity and forging forward
.
Intellectual and Developmental Disabilities
,
51
,
423
427
.
Houseworth,
J.,
Stancliffe,
R. J.,
&
Tichá,
R.
(2018)
.
Association of state-level and individual-level factors with choice making of individuals with intellectual and developmental disabilities
.
Research in Developmental Disabilities
,
83
,
77
90
.
Human Services Research Institute & The National Association of State Directors of Developmental Disabilities Services
.
(2018)
.
2016 staff stability report: ACS final report
.
Lakin,
K. C.,
&
Bruininks,
R. H.
(1981)
.
Occupational stability of direct-care staff of residential facilities for mentally retarded people
.
University of Minnesota
,
Center on Residential and Community Services
.
Larson,
S. A.,
Eschenbacher,
H. J.,
Anderson,
L. L.,
Taylor,
B.,
Pettingell,
S.,
Hewitt,
A.,
Sowers,
M.,
&
Bourne,
M.L.
(2018)
.
In-home and residential long-term supports and services for persons with intellectual or developmental disabilities: Status and trends through 2016
.
University of Minnesota, Research and Training Center on Community
.
Larson,
S. A.,
Hewitt,
A. S.,
&
Lakin,
K.C.
(2004)
.
Multiperspective analysis of workforce challenges and their effects on consumer and family quality of life
.
American Journal on Mental Retardation
,
109
(6)
,
481
500
.
Larson,
S.,
&
Sedlezsky,
L.
(2015)
.
Assessing retention outcomes
.
Research Training Center on Community Living
,
Institute on Community Integration, University of Minnesota
.
Larson,
S. A.,
Tolbize,
M. S. A.,
Kim,
O.,
&
York,
B.
(2016)
.
Direct support professional turnover costs in small group homes: A case study
.
University of Minnesota, Research and Training Center on Community Living
.
Lerman,
R. I.,
Eyster,
L.,
&
Kuehn,
D.
(2014)
.
Can we upgrade low-skill, low-wage occupations? The case of apprenticeships in the long-term care occupations
.
Journal of Women, Politics & Policy
,
35
(2)
,
110
132
.
Nadeau,
C. A.,
&
Glasmeier,
A. K.
(2018)
.
Living wage calculator: User's guide/technical notes
.
Massachusetts Institute of Technology
.
National Academies of Sciences, Engineering, and Medicine
.
(2016)
.
Strengthening the workforce to support community living and participation for older adults and individuals with disabilities: Proceedings of a workshop
.
The National Academies Press
.
National Alliance on Direct Support Professionals
.
(2016)
.
Direct support professional competency areas: The foundation of direct support practice
.
Author
.
National Core Indicators
.
(2014)
.
National core indicators 2014 staff stability survey report
.
National Core Indicators
.
(2018)
.
National core indicators 2017 staff stability survey report
.
National Federation of Independent Business (NFIB) v. Sebelius, 567 U.S. 519
(2012)
Oklahoma Department of Human Services, Developmental Disabilities Service Division
.
(2000)
.
Direct support staff turnover reduction plan: Annual report to residential providers
.
Oklahoma Department of Human Services
,
Developmental Disabilities Service Division
.
Paraprofessional Healthcare Institute
.
(2015)
.
Paying the price: How poverty wages undermine home care in America
.
Author
.
Paraprofessional Healthcare Institute Policy Works
.
(2014)
.
Facts 1: Occupational projections for direct-care workers, 2012-2022
.
Author
.
Patient Protection and Affordable Care Act, 42 U.S.C. § 18001 et seq
.
(2010)
.
Pettingell,
S.,
Kramme,
J.,
&
Hewitt,
A.
(2019)
.
Minnesota direct support workforce 2018 final report
.
St. Paul
:
Minnesota department of human services
.
Powers,
E. T.
&
Powers,
N. J.
(2010)
.
Causes of caregiver turnover and the potential effectiveness of wage subsidies for solving the long-term care workforce ‘crisis.' The B.E. Journal of Economic Analysis & Policy, 10(1)
.
President's Committee for People with Intellectual Disabilities
(2017)
.
Report to the President 2017 America's direct support workforce crisis: Effects on people with intellectual disabilities, families, communities and the U.S economy
.
The Administration for Community Living
.
Raudenbush,
S. W.,
&
Bryk,
A. S.
(2002)
.
Hierarchical linear models: Applications and data analysis methods (Vol. 1)
.
Sage
.
Raudenbush,
S. W.,
Bryk,
A. S,
&
Congdon,
R.
(2013)
.
HLM 7.01 for Windows [Computer software]
.
Scientific Software International
,
Inc
.
Skirrow,
P.,
&
Hatton,
C.
(2007)
.
‘Burnout' amongst direct care workers in services for adults with intellectual disabilities: A systematic review of research findings and initial normative data
.
Journal of Applied Research in Intellectual Disabilities
,
20
(2)
,
131
144
.
Test,
D.,
Solow,
J.,
&
Flowers,
C.
(1999)
.
North Carolina direct support professionals study: Final report
.
University of North Carolina at Charlotte
.
Test,
D. W.,
Flowers,
C.,
Hewitt,
A.,
&
Solow,
J.
(2003)
.
Statewide study of the direct support staff workforce
.
Mental Retardation
,
41
(4)
,
276
285
.
U.S. Census Bureau
.
(2016)
.
American Community Survey 1-Year Estimates
.
U.S. Department of Health and Human Services
.
(2019)
.
U.S. federal poverty guidelines used to determine financial eligibility for certain federal programs
.
Vassos,
M. V.,
&
Nankervis,
K. L.
(2012)
.
Investigating the importance of various individual, interpersonal, organizational and demographic variables when predicting job burnout in disability support workers
.
Research in Developmental Disabilities
,
33
(6)
,
1780
1791
.
Wiener,
J. M.,
Squillace,
M. R.,
Anderson,
W. L.,
Khatutski,
G.
(2009)
.
Why do they stay? Job tenure among certified nursing assistants in nursing homes
.
The Gerontologist
,
49
(2)
,
198
201
.

Author notes

Development of this article was supported by Grant #90RTCP0003 to the Research and Training Center for Community Living from the National Institute on Disability Independent Living and Rehabilitation Research, U.S. Department of Health and Human Services.