GWMI Health Workforce White Paper #2

Welcome to interactive presentation, created with Publuu. Enjoy the reading!

October 2016

Does ACO Adoption Change

the Health Workforce

Configuration in U.S.

Hospitals?

AUTHORS:

Avi Dor, PhD

Patricia Pittman, PhD

Clese Erikson, MPAff

Roberto Delhy, MA

Xinxin Han, MS

Mullan Institute Health

Workforce White Paper No. 2

Prepared By

The George Washington University

Fitzhugh Mullan Institute for Health Workforce Equity

Questions

For questions regarding this report, please contact

Patricia Pittman at ppittman@gwu.edu.

Suggested Citation

Dor A, Pittman P, Erikson C, Delhy R, Han X. Does ACO Adoption Change the Health

Workforce Configuration in U.S. Hospitals?. Washington, DC: Fitzhugh Mullan Institute for

Health Workforce Equity, George Washington University; 2016.

Funding

This white paper was supported by the Bureau of Health Workforce (BHW), National

Center for Health Workforce Analysis (NCHWA), Health Resources and Services

Administration (HRSA) of the U.S. Department of Health and Human Services (HHS) as

part of an award totaling $450,000, with zero percent financed with non-governmental

sources. The contents are those of the author[s] and do not necessarily represent the

official views of, nor an endorsement by HRSA, HHS, or the U.S. Government.

1

October 1, 2016

f

Does ACO Adoption Change the Health Workforce

Configuration in U.S. Hospitals?

Table of Contents

INTRODUCTION ............................................................................................................................................. 2

JOBS IN PREMIER DATA ................................................................................................................................. 3

JOBS IN AHA DATA ........................................................................................................................................ 4

CONCLUSIONS ............................................................................................................................................... 6

REFERENCES .................................................................................................................................................. 9

List of Tables

Table 1: Hospital staffing by ACO status, 2014 ........................................................................................... 11

Table 2: Distribution of Shared Savings Programs in 2013 and 2014 ......................................................... 12

Table 3: Changes in Hospital staffing by ACO status, 2014......................................................................... 13

2

October 1, 2016

f

INTRODUCTION

An Accountable Care Organization (ACO) refers to a group of physician and other healthcare

providers and suppliers of services, notably hospitals that form a collaborative network in order to reduce

costs while improving quality of inpatient care, and thereby meet contractual requirements and targets

set by third-party payers. Although the number of ACOs nationwide has been growing rapidly since they

were first recognized in Medicare regulations in 2011, little is known about the way delivery systems

adjusted to the change, and specifically about how those changes affect the health care workforce in

hospitals).1 The purpose of this report is to provide a descriptive analysis of workforce differences

between hospitals that participate in an ACO and those that do not.

The most well-known type of ACO is the Medicare Shared Risk Program (MSPP). However, recent

surveys suggest that commercial ACO contracts are rising rapidly. It is estimated that by the end of 2015

the number of lives covered by commercial ACO contracts was more than double the number of lives in

Medicare ACOs – 17.2 and 8.3 million, respectively.2

Nearly all Medicare ACOs opted for one-sided risk contracts offered by CMSi, whereby they would

share savings with CMS if costs of the patient pool are below some threshold payment level, up to 50% of

the spending difference. In addition, in order to be eligible to share in any savings generated, an ACO must

meet the established quality performance standard that corresponds to its performance year.3

One idea put forth by experts was that to ensure better outcomes ACO hospitals would likely

increase primary care clinicians, including nursing care staff.4 The premises was the belief that an increase

in staffing would help ACOs to adjust to the new regulations and standards arising from the

implementation of the ACA, and would help them gain legitimacy and credibility among payers and

patients.4

Extending beyond the immediate transition of care, hospital engagement in care management of

complex patients is associated with reduced readmissions.5 Given that a relatively small portion of the

population accounts for the majority of health care expenditures, identifying high risk patients, including

patients with social service or behavioral health needs, and focusing care on that population is an

important cost containment strategy.6,7 This can take the form of developing disease registries to track

patients with one or more chronic diseases8 to complex algorithms to prospectively identify patients using

claims data and other patient information.9 The personnel handling these activities may be listed as data

analysts, or they may have other titles such as medical assistants.1

While all of these activities would suggest an increase in nurse staff, as well as, possibly, care

coordinators, data analysists, pharmacists, and others, the manner in which payments are determined

may actually provide a disincentive for hospitals to spend on labor once they begin the program. Douven

and colleagues point out that benchmark calculations of payment include the last three years of providers’

spending, but that it is the most recent year that carries the greatest weight.10 Thus, the incentives to

increase spending are strongest in the last year prior to ACO adoption. They argue that providers that

incur the greatest costs during the year before entering or renewing an ACO arrangement are actually

rewarded, while providers with the lowest cost during the last year are in fact penalized.22 Under these

conditions, it is possible that joining an ACO would have the effect of reducing previously inflated staff

levels in high-cost and high-benchmark ACOs.

Given the conflicting effects that ACO adoption may have on nursing and other staff hiring,

coupled with the fact that incentives in commercial ACOs are not known, we believe that an exploratory

analysis of staffing in ACO programs is warranted.

i Only 2% of ACOs opted for the two-sided risk model, where they split both losses and gains with CMS

3

October 1, 2016

f

To conduct this preliminary analysis, we used two different data sources as follows:

First, we used an operations database maintained by Premier to examine a set of jobs that

might be related to care coordination to see if ACO hospitals differed from non ACO hospitals

in their use of each type of personnel. This analysis was limited to a cross-sectional comparison

for 2014, the only year for which Premier had tagged ACOs.

Second, we used the American Hospital Association’s (AHA) Survey of Care Systems and

Payment to identify ACOs, and the AHA’s regular annual survey to examine nurse staffing

ratios in the two types of hospitals in 2013 and in 2014, the two years for which data is now

available.

The methods and findings for each of these analyses is further detailed below in each section. We

conclude with a section that lays out the next steps in this program of research and describes what we

have learned about the relative strengths of our data sources for workforce research.

JOBS IN PREMIER DATA

Methods & Data

To examine the extent to which becoming an ACO, and the increased use of care coordination as

documented above, might lead to new jobs or changes in the configuration of staff in hospitals, we

compared hospitals in 2014 that were ACOs to those that were not. We used an operational database

maintained by Premier Inc. that tracks labor hours, hospital units, and facility characteristics.

A total of 317 unique hospitals were included in 2014 data. The ACO flag variable was linked by

Premier from American Hospital Association (AHA) 2014 Annual Survey. The sample included 135

hospitals that were ACOs, and 182 hospitals were non-ACOs.

Measures

Based on a review of job titles in the Premier data, we identified 16 jobs that might be affected

by ACO status. These jobs included advanced practical registered nurse (APRN), physician assistant (PA),

registered nurse (RN), licensed practical nurse (LPN), unlicensed assistive personnel (UAP), nurse assistive

personnel (NAP), case manager and case management assistant, managed care coordinator, risk

management, patient educator, social worker, medical social worker, clinical social worker, pharmacists,

and pharmacy technician.

Hospital staffing was measured by annual average number of hours worked (including only regular

and overtime hours) by each type of workforce examined. The labor hours were also adjusted by case-

mix index adjusted total patient days, a similar measure used in previous nursing studies. Compared to

full- time equivalent workers, this measure allowed us to capture the impacts of absences from work, as

well as overtime hours.

We then conducted a cross-sectional analysis of ACOs staffing and used t-tests to compare staffing

levels between ACOs and non-ACOs in 2014.

Results

As presented in Table 1, in 2014, ACOs and non ACO hospitals used similar levels of PA, UAP, NAP

and APRN staffing (0.073 vs. 0.053, p=0.216; 1.053 vs. 1.142, p=0.242; 0.989 vs. 1.105, p=0.117; 0.167 vs.

0.169, p=0.938). However, ACOs used significantly lower level of RN but higher level of LPN staffing than

non-ACOs in 2014 (3.891 vs. 4.772, p=0.000; 0.290 vs. 0.216, p=0.029).

In 2014, ACOs and non ACOs used similar levels of case manager and case management assistant

staffing (0.122 vs 0.143, p=0.088). Likewise, ACOs used similar levels of managed care coordinator, risk

management, and patient educator staffing as non-ACOs (0.011 vs. 0.016, p=0.364; 0.013 vs. 0.016,

p=0.467; 0.018 vs. 0.019, p=0.692; 0.018 vs. 0.019, p=0.692). However, ACOs used lower levels of social

4

October 1, 2016

f

workers, pharmacists, and pharmacy technician staffing than non-ACOs (0.076 vs. 0.097, p=0.028; 0.138

vs. 0.188, p=0.000; 0.169 vs. 0.198, p=0.005).

Discussion

We are cautious in our interpretation of these results, given that the Premier data and the AHA

ACO flag are both convenience sample. However, it is notable that there are no significant increases in

hours for care coordinators, case managers, patient educators, or risk managers, as might have been

expected. It is also notable that RN and UAP hours are lower in ACO hospitals than in non ACO hospitals,

as are pharmacy and social worker hours.

These differences are the opposite of what we might have expected, given the evidence that

higher RN staffing levels improve outcomes11,12,13, and the increased interest in pharmacists for

medication management14 and social workers to help manage patients with challenges in the realm of

social determinants15. They may suggest that ACO hospitals are engaged in cost containment strategies

that include constraining staff growth.

Further analyses that includes additional years and multivariate regressions that can control for

facility and regional characteristics are needed and will be conducted during the fall of 2016.

JOBS IN AHA DATA

The purpose of this second section of the report is to continue to explore workforce differences

in ACO hospitals using the AHA Annual Survey Database, and the AHA Survey of Care Systems for the years

2013 and 2014. Henceforth we will refer to these as the 2013 and 2014 AHA and ACO surveys respectively.

From the AHA and ACO surveys, our 2013 data included information for 1,795 hospitals. Out of

these 1,795 hospitals, we excluded 358 government hospitals, leaving us with 1,250 non- governmental,

not-for-profit hospitals and 187 investor-owned for-profit hospitals. We excluded government hospitals

to date the literature does not address the role that federal/local hospitals in ACO. In later analysis we

aim to include/compare in our analyses federal/local hospitals. Out of the resulting 1,437 private

hospitals, therefore, 286 had “established or were part of an ACO”, while 853 of them did not have any

type of ACO arrangementii.

For the year 2014, both the AHA and ACO surveys included information for 1,697 hospitals. Out

of these 1,697 hospitals, we excluded 317 government hospitals, leaving us with 1,239 non-governmental,

not- for-profit hospitals, and 141 investor-owned for-profit hospitals. Out of the resulting 1,380 private

hospitals, 279 had “established or were part of an ACO”, and 526 did not have any type of ACO

arrangement.

After merging our 2013 and 2014 information of private hospitals with ACO information for both

years, our final dataset was composed by 403 hospitalsiii. Out of those, 117 hospitals were part of an ACO

in both years 2013 and 2014 (group #1 – always ACO), 268 hospitals were not part of an ACO in either

2013 or 2014 (group #2 – never ACO), 63 hospitals were not part of an ACO in 2013 but joined an ACO in

2014 (group #3) and finally 25 hospitals were part of an ACO in 2013, but reported having left their ACO

arrangements by 2014 (group #4).

With regards to the rural/urban distribution of the hospitals described above, using US Census

defined Core Based Statistical Areas, we determined that the number of rural hospitals was very small in

all groups except the never-ACO group. It included 8 rural hospitals in group #1, 104 rural hospitals in

group #2, 8 rural hospitals in group #3 and 6 rural hospitals in group #4.

ii For 298 of these hospitals, ACO status information was missing

iii Out of 1,380 non-governmental hospitals present in both years, 575 of them had missing ACO status information for either year.

5

October 1, 2016

f

In regard to the variety of shared savings programs pursued by ACO hospitals, the three most

common shared savings arrangements in 2013 and 2014 were the following: commercial payer

partnerships, Medicare shared savings programs, and joint Medicare shared savings programs together

with commercial payer partnerships (see Table 2).

As observed in Table 2, the distribution of ACO participating hospitals in different shared savings

programs is complex and dynamic. A significant number of hospitals entered and exited the existing

shared savings programs just within the two-year observed period. Additionally, hospitals may have

multiple shared savings programs which can also change from year to year.

For the workforce analysis, we consider only hospitals in groups #1 (always ACO) and #2 (never

ACO) as described above. We excluded the two groups of hospitals were ACO membership status changed

between 2013 and 2014, because of small sample sizes which would prevent us from obtaining

meaningful statistical comparisons.

We focus on the number of staff hours per adjusted patient day. Total patient days are adjusted

by the hospital level proportion of inpatient and outpatient revenue, per AHA’s methodology, and we

then also adjusted for CMS’ yearly case mixed indexes (CMI).

We use a modified measure of full-time equivalent employment developed by Spetz and

colleagues.16 They assume that productive hours per year are fewer than 2,080 and instead use 1,768

hours per year; this is equivalent to an 85% productive level over 52 weeks per year at 40 hours per week.

Based on staffing data available from the AHA, we focused this analysis on various levels of nursing

staff: registered nurses (RNs), licensed practical nurses (LPNs), nursing assistive personnel (NAP) (the term

used in the AHA survey), and advanced practice nurses (APNs). As discussed in the introduction, there is

extensive research demonstrating that higher RN staffing levels result in better outcomes. LPN and APN

staffing levels may affect RN workload and therefore may also affect outcomes. APNs do various types of

tasks in hospitals, and we know of no research suggesting APN staffing levels bear a relationship with

outcomes. We include them simply as an additional point of interest that could be explored in future

research.

Thus, for each type of nurse we use the following formula in order to calculate the number of

hours per CMI adjusted patient day:

After calculating the average number of nursing hours for RNs, LPNs, NAPs and APNs, we then

proceed to compare the absolute difference, as well as the percentage change across occupations among

ACO and non-ACO hospitals between 2013 and 2014.

Results

As presented in Table 2, when comparing the baseline (2013) staffing levels for ACOs and non

ACOs, we find that non ACOs have higher levels of LPNs and APNs. Given the higher proportion of rural

hospitals in the non ACO group, this is not surprising. We know that nationwide employment of LPNs in

hospitals is falling, although it is most dramatic in urban areas. We also know that employment of APNs is

rising across the country, but especially in rural areas.17

In comparing changes in the two groups from 2013 to 2014, we find that the average number of

RN and APN hours has fallen in the always-ACO group and that the changes are statistically significant.

This finding was surprising, as discussed below, given the idea that ACOs may be more advanced in the

organization of care than non ACOs.

6

October 1, 2016

f

We also find that LPN and NAP hours fell in both groups, while APN hours increased significantly

in both groups, especially in the always-ACO group.

Discussion

We are again cautious not to interpret our results as a definitive study of labor patterns among

ACO/non-ACO hospitals, given the potential for sample selection bias and the fact that this is simply a

descriptive analysis. Nevertheless, findings relating to the small drop in RN and the large increase in APN

hours per patient day in ACOs echo our findings using Premier data for just one year (2014).

CONCLUSIONS

We were surprised that not only did there not appear to be major ACO related differences in the

workforce either cross sectionally or when comparing changes across two years in the “always ACO” and

“never ACO” groups. Indeed the early signs of directionality of change among those with ACO status are

the opposite of what we would have anticipated. These two descriptive analyses suggests the need for

further research with longer time series (more data), as well as, adjustments that take into account other

important characteristics of hospitals.

ACO’s are required to report quality outcomes and their payment is linked to results. Indeed, we

know that ACOs that participated in the Medicare Shared Savings program in both 2013 and 2014

improved on 27 of 33 measures.18

Given, increased interest in pharmacists for medication management19 and social workers to help

manage patients with challenges in the realm of social determinants20 makes the lower staffing levels of

these groups of professionals in ACO hospitals unforeseen.

Even more surprising was the reduction in RN and APN hours among ACO hospitals, as compared

to non ACO hospitals, given that there is such a strong body of research suggesting higher nurse staffing

results in better outcomes. Moreover, nurse organizations, especially labor unions, have been advocating

for various types of mandatory and voluntary nurse staffing laws. Indeed, California and Massachusetts

now have mandatory laws, and 14 other states have either public reporting or staffing committee

requirements that aim to push hospitals towards higher nurse staffing levels.21 At the same time, recent

research conducted by this team (forthcoming in HSR) using Premier data also shows a nationwide decline

in RN and nurse support staff hours, suggesting that something important may be occurring.

With regard to reduced or constrained staffing, possible explanations to be explored in further

research include the following:

A suggested by Douven and colleagues, the existing ACO payment formula may lead hospitals

to reduce spending just after joining ACOs. This may be due simply to the savings incentives,

or it may be linked to the payment calculations which give the greatest weight to the year

prior to ACO adoption.

Hospitals in ACOs are expanding their market power and, therefore, may be less concerned

about public reputation regarding nurse staffing, and perhaps even their nurse related quality

outcomes;

ACO hospitals are identifying ways to reduce labor costs without affecting outcomes;

The reduction in RN staff could be related to the retirement of older nurses, and hospitals in

certain regions may be having difficulty replacing them.

It is possible that ACO hospitals are redeploying some RNs and support staff to other settings

with partner organizations and they are therefore not captured in either data source.

The next steps in our research will be to use a longer longitudinal data set and multivariate

regression analyses to control for a variety of facility and regional characteristics to ensure that our

preliminary findings are robust. Among the control variables that should clearly be included in our next

7

October 1, 2016

f

analysis are rurality, local nurse supply, hospital market share, types of shared savings programs, and

percent of patients covered under shared savings.

Reflections on the Strengths and Weaknesses of our Data Sources

One of the important objectives of this phase of our research is to fully comprehend these new

data sources for workforce research. In the paragraphs that follow, we reflect on the relative strengths

and weaknesses of Premier and AHA data sources that we have identified to date.

Premier Data

The Premier operational database captures information on more than 500 Premier membership

hospitals that cover healthcare systems in all 50 states and the District of Columbia. It includes basic

facility characteristics, department codes and descriptions, job titles and descriptions, and staffing

information such as labor hours, expenses, and skill-mix category. GW has purchased 2010-2014 and 2015

will be provided to us soon. Premier collects clinical, financial, pharmacy, supply chain, and operational

data from its member hospitals on a daily, biweekly, monthly, or quarterly basis. The data provide a

unique opportunity to track hospital-based workforce, and in this instance it allows us to identify a variety

of job titles across different hospital departments and specialty areas.

Hospital staffing was measured by calculating the annual number of total worked hours (including

regular and overtime hours) for each of the selected jobs, adjusted by case-mix index adjusted total

patient days. Compared to full-time equivalent workers, this labor hour measure allows us to capture the

impacts of absences from work, and thus may reflect the actual hours that workers spend on assisting

clinical tasks. Premier was able to link the ACO flag from AHA for us (we are not provided access to

provider identifying numbers), allowing us to compare the differences in care coordination related

workforce in hospitals.

The weakness of the Premier data is the small sample size, which may affect the statistical power

to detect statistical differences. That is to say, while some of our results did not have significant

differences, the magnitude of the estimates still provide practical implications. In addition, Premier’s

member hospitals are essentially a convenience sample of all U.S. hospitals and thus may not necessarily

be representative of all U.S. hospitals. However, Premier’s hospitals characteristics are still similar to the

characteristics of U.S. community hospitals as reported by other national hospital databasesiv suggesting

that our findings are likely to reflect hospital staffing trends nationwide. Lastly, the ACO flag linked from

AHA indicates ACOs, being part of ACOs, however, it also indicates hospitals that are not ACOs but are

actively thinking to become an ACO as the ACO flag. This may affect the accuracy of our estimates. Future

study may need to exclude these hospitals out.

American Hospital Association (AHA) data

The AHA Annual Survey Database (ASDB) covers all U.S. community health hospitals, and the

response rates are high in each year. GW has purchased 2009-2014. This annual survey contains over

6,300 hospitals and almost 1,000 fields of information in the following categories: organization structure,

facility characteristics, inpatient and outpatient utilization, staffing, and geographic indicators. Due to its

reliability across time, the data are used for a variety of purposes. They are seldom used, however, to

examine the hospital workforce. Two previous works we found are using AHA data as regression

controls.22,23 The AHA annual survey provides data on number of FTE for a smaller number of jobs than

iv The proportion of teaching hospitals and average occupancy rate in our dataset is comparable to the national average, while our sample

consists of a larger portion of not-for-profit, urban, and system-affiliated hospitals and hospitals with more staffed beds and admissions as

compared to the national sample from the American Hospital Association Annual Survey and Healthcare Cost and Utilization Project data.

8

October 1, 2016

f

Premier. These include physician, dentists, medical residents, nurses, nurse assistive personnel,

technicians, pharmacists, therapists, other clinical professionals, and support personnel.

The AHA Survey of Care Systems and Payment was established in 2013 to understand payment

arrangements. GW purchased 2013-2015. The most important component of this survey is its focus on

Accountable Care Organization (ACO) participation and structure. The survey identifies which hospitals

are participating in ACOs, or are actively thinking to become an ACO. Using this information, we are able

to link the AHA annual data workforce variable to the ACO flag created from the ACO survey and examine

staffing variation by hospital ACO status.

In 2013, the survey was sent to all registered community hospitals (4,999) and received 1,517

responses from the field. Of these, 309 hospitals indicated they are part of an ACO. The respondent profile

for the overall survey was broadly representative of the universe of U.S. hospitals, as indicated by the

AHA.24

Both of the AHA surveys have limitations. First, the AHA ASDB does not distinguish between

outpatient and inpatient registered nurses (RNs). Second, surveyed hospitals may use different definitions

to calculate the number of FTE workers. Thirdly, when we convert the number of FTE nurses to annual

nursing hours, we use a formula previously used by Spetz and colleagues in which they assume that one

FTE nurse would work for 1,768 hours per year. Under this formula, potential work hours equal to 52

weeks per year at 40 hours per week and actual productive hours equal to 85% of potential hours.71 This

method yields similar results to Premier data.

Lastly, the ACO surveys’ response rates are relatively low, resulting in the possibility of selection

bias. In other words, hospitals that chose to answer the survey may differ from hospitals that did not

answer it in ways that we do not understand. Future research will need to account for this issue.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

Made with Publuu - flipbook maker