Is Technology Eating Nurses? --- Staffing Decisions in Nursing Homes
Abstract
Over the past ten years, many healthcare organizations have made significant
investments in automating their clinical operations, mostly through the introduction of advanced
information systems. Yet, the impact of these investments on staffing decisions and on optimally
managing delivery costs—all while trying to enhance the quality of care - is still not well
understood. In this paper, we study the effect of IT-enabled automation on staffing decisions in
healthcare facilities. Using unique nursing home IT data from 2006 to 2012, we find that the
staffing level decreases by 5.8% in high-end nursing homes but increases by 7.6% in low-end
homes after the adoption of automation technology. The outcomes can be explained by the
interplay of two competing effects of automation: the substitution of technology for labor and the
leveraging of competitiveness via technology. We also find that increased automation reduces
resident complaints about process quality by 36.8% and decreases the admissions of less
profitable residents by 14.7% on average. These observations are consistent with the predictions
of a staffing model that incorporates technology adoption and vertical differentiation. Overall,
these findings suggest that the impact of automation technology on staffing decisions depends
crucially on a facility’s strategic position in the local marketplace.
Key Words: process quality, staffing, labor, automation technology, nursing homes, vertical
differentiation
1. Introduction
Will improved technology provide and produce enough work? In a recent Wall Street
Journal article, former U.S. Treasury Secretary Lawrence Summer voiced just such a concern.1
Microsoft co-founder Bill Gates warned in 2014 that automation threatens all manners of
workers, from drivers to waiters to nurses. 2 Similarly, but in a more optimistic tone, Marc
Andreessen, 3 a well-known entrepreneur and software engineer, said in an interview that
“software will eat the world” to express his view that information technology will revolutionize
1
http://www.wsj.com/articles/lawrence-h-summers-on-the-economic-challenge-of-the-future-jobs-1404762501
http://www.wsj.com/articles/what-clever-robots-mean-for-jobs-1424835002
3
http://www.wired.com/2012/04/ff_andreessen/5/.
2
1
all sectors of the economy, from retail to real estate to health care, and will replace old jobs with
new jobs along the way.
Advances in information technology are changing healthcare delivery by bringing
digitization and automation into the industry. As more and more healthcare providers adopt
technologies such as electronic medical record and computerized physician order entry (CPOE),
the administration and delivery of care become streamlined and efficient (Zlabek et al, 2011, Lee
et al., 2013). As a result, policy makers and medical providers, especially nurses, want to know:
is technology “eating” nurses, the basic labor force in the healthcare industry?
In this study, we aim to address the technology-nurse relation from a strategic
positioning perspective. Although the relation between technology and nurse employment has
generated considerable interests among the public, the literature investigating the causal effect of
IT-enabled automation on staffing decisions in individual healthcare facilities is relatively sparse
for two reasons. First, it is difficult to find an ideal setting where tasks taken by different types of
labor are well defined and detailed staffing data at the facility level are readily available. Second,
the adoption of automation technology is not exogenous. In the technology adoption literature,
only a limited number of studies carefully consider this type of endogeneity issue (e.g. Miller
and Tucker, 2011).
To address the first issue, we choose nursing homes as the study subjects because quality
of care in a nursing home is mainly determined by nurses on a daily basis. IT-enabled
automation may help reduce nurse tasks, increase the utilization of nurse time and improve the
working environment for nurses. Interestingly, adoption of IT-enabled automation is largely
lagged behind in nursing homes compared to other healthcare settings such as hospitals and
emergency care. According to the Health Information Systems Society (HIMSS) data, more than
80% of hospitals adopted the CPOE while less than 40% of nursing homes implemented it by
2011. One major barrier for the adoption comes from nurses, the major input for providing care
in nursing homes. Being concerned about possible unemployment, nurses hesitate to push for the
health IT adoption despite potential benefits from automation. Therefore, it is important for the
nursing home industry to better understand the implications of IT adoption on nurse employment.
In addition, the Online Survey Certificate and Reporting Database (OSCAR) provides the full
time equivalents (FTEs) of specific nurse types, which allows us to study many staffing-related
questions.
2
We take an instrumental variable (IV) approach to address the endogeneity concern. We
use the adoption of CPOE, an advanced information system, to measure the level of IT-enabled
automation in nursing homes. CPOE can help healthcare providers to streamline operations via
automation, reduce medication errors, and improve patient safety (Galanter et al, 2005, Zlabek et
al, 2011). The IV we construct comprises the hospital CPOE adoption rates in the local market.
The idea is that a nursing home’s decision on technology adoption responds to the adoption
decisions of local hospitals. However, the upstream hospital adoptions are unlikely to directly
affect quality and staffing decisions of an individual nursing home in the downstream. We
construct this instrumental variable following Miller and Tucker (2009) which shows that greater
adoption of information technology by other hospitals should lead to greater network benefits to
a hospital from its own adoption of information technology.
Using the OSCAR data and the HIMSS data from 2006 to 2012, we find that the effect
of automation on staffing decision depends crucially on a nursing home’s vertical position in the
local market. Our results show that the licensed nurse (LN) staffing level increases on average
7.6% in low-end nursing homes but decreases by 5.8% in high-end homes after the CPOE is
adopted. The outcomes can be explained by the interplay of two competing effects of IT-enabled
automation. On one hand, automation increases the marginal benefit of quality improvement
from increased staffing. This complementarity effect helps a nursing home improve its
competitiveness in its local market. On the other hand, the marginal effect of quality on revenue
diminishes as quality improves. Given that the marginal cost of staffing is constant, an increase
in automation may actually lead to the substitution of technology for labor, which we call the
substitution effect. The complementarity effect dominates the substitution effect in low-end
nursing homes, but is dominated by the latter in high-end ones.
Further, we find that an increase in automation leads to an increase in the process
quality (measured by the number of resident complaints). The results show that resident
complaints decrease by 36.8% on average after a nursing home implements the CPOE, all else
being equal. As a result, although CPOE adoption does not change the number of total
admissions possibly due to capacity limits, it does result in a 14.7% decrease in the admissions of
Medicaid patients, the least profitable type, regardless of a nursing home’s vertical position.
These results echo with the current industry trends that nursing homes strive for quality
3
improvement as they chase lucrative patients. 4 Moreover, all these empirical findings are
consistent with the predictions of a staffing model that incorporates technology adoption and
vertical differentiation.
2. Industry Background and Literature Review
A nursing home is a place for the elderly or the disabled who do not need to be in a
hospital but still require different levels of assistance in daily living activities and/or medical
care from professional nurses (Lu and Lu, 2014). Due to the aging of the baby boomer
generation, approximately $ 111 billion was spent on nursing home care in the United States in
2011, up from $ 92 billion in 2006.5 The entire nursing home industry is very competitive.
According to the OSCAR data, on average, 13.7 nursing home companies located within a five
mile radius. Compared to hospitals, nursing homes provide relatively homogeneous services
and quality is mainly determined by the nurses (Norton, 2000). They thus provide an ideal
setting to investigate the relation among automation technology, staffing input and process
quality.
2.1 Quality Mix and Vertical Differentiation
Nursing home services are mainly paid in one of the three ways: by Medicare, by
Medicaid, or by a private party.6 Short-term services are generally paid by Medicare or private
insurers, and long-term services are largely paid by Medicaid. Currently, Medicare patients
generate the highest per-resident revenue and profit margins (average revenue of around $500
per resident per day in 2015)7 while Medicaid residents generate the lowest (average revenue
under $194 in 2015)8. Hence, nursing homes are often assessed in terms of “quality mix”, an
industry jargon defined as the percentage of revenues from sources other than Medicaid (Credit
Suisse Equity Research, 2001). Since prices for government patients are set by public payers and
4
http://omnifeed.com/article/www.nytimes.com/2015/04/15/business/as-nursing-homes-chase-lucrative-patients-qualityof-care-is-said-to-lag.html
5
http://www.ltlmagazine.com/news-item/ltc-industry-generates-259-billion-revenue-during-2011
6
The distribution of the three types of patients in 2006 was as follows: Medicaid (64.8%), Medicare (13.4%) and Privatepaying (21.0%).
7
To qualify, patients must spend at least 3 full days in a hospital before their admission into a nursing home, and they
must show a need for skilled nursing care. In addition, the transfer from a hospital must occur within a certain time.
Medicare pays for the first 20 days at full cost, and the difference between $114 per day and the actual cost for up to
another 80 days (The BBA of 1997).
8
These daily rates are taken from the quarterly report filed by Genesis Healthcare on May 8, 2015.
4
nursing homes are not allowed to discriminate residents based on their payer types (Grabowski et
al, 2008), competition for the higher-margin Medicare and private-paying residents is based
primarily on the differential quality of care. As the Medicare budget has been trimmed for
hospital care and assisted living continues to encroach on the private market, the quality of care
provided by a nursing home is becoming an increasingly important factor affecting its margin.
Vertical differentiation in service quality is an important feature in the nursing home
industry, which is directly reflected by the star rating system shown at the federal Nursing Home
Compare (NHC). For example, one of the largest nursing home chains, Kindred, mainly targets
Medicaid patients and provides relatively low quality of care. Hence, most of its units are lowend nursing homes rated three stars. Corresponding to this vertical position strategy, Kindred
seldom uses a brand name for its individual units so as to minimize the negative externality
across sibling units (Brickley et al, 2014). By contrast, HCR Manor Care is a premium company
in this industry with many high-end nursing homes. It owns a great brand name Manor Care
with most of its units rated five stars.
Besides, the NHC also publicly released staffing ratios to help consumers identify high
quality nursing homes. Konetzka et al (2009) and Lin (2014) show that high staffing ratios are
positively associated with high quality in nursing homes. Naturally, nurse input thus serves as a
critical proxy for quality. According to the 2006 staffing information released at the NHC, the
mean licensed nurse (including registered nurse and licensed practical nurse) hours per resident
day is 1.66 HPRD, and its dispersion between 1 percentile and 99 percentile is 4.59 HPRD,
suggesting that nursing homes offer services at different quality levels and target different
consumer segments.
2.2 Literature Review
Our work is related to the economics and management literature on labor inputs and
technology choices. The consensus view in this literature is that skilled labor is complementary
to technology (e.g., Griliches 1969; Berman, Bound, and Griliches 1994; Autor, Katz, and
Krueger 1998; Krusell et al. 2000; Acemoglu 2002). By contrast, Acemoglu and Finkelstein
(2008) present evidence that the introduction of prospective payment system (PPS), which
increases the relative price of labor, is associated with an increase in technology adoption and a
5
decline in labor inputs, suggesting that there is a relatively high degree of substitution between
technology and labor.
Most relevant to our study is a small set of papers examining the substitution of IT
resources for labor. Dewan and Min (1997) formulate a CES-translog production function with
IT capital, non-IT capital, and labor as inputs, and annual value added by a firm as the output.
They find that IT capital is a net substitute for ordinary labor in all sectors of the economy. Lin
and Shao (2006) study pairwise substitutability among IT stock, traditional capital, and
traditional labor. They find the three factors are not pairwise substitutable and it is not possible to
use IT stock to replace labor entirely. More recently, Chwelos et al. (2009) find that the
increasing factor share of IT comes at the expense of labor through labor substitution. Furukawa
et al. (2010) estimate a translog production function and find that IT has a substitutability effect
on low-skilled labor and an even stronger complementarity effect on higher-skilled workers. All
these papers use an abstract production function to model the mechanism through which IT
affects the output. Our paper contributes to this line of literature in three ways. First, we
explicitly model how technology affects a firm’s output by analyzing the firm’s optimal labor
decision after investing in technology. Second, we incorporate the vertical position of a firm
within the industry to study the impact of IT adoption on a firm’s strategic staffing decisions.
Third, instead of using a production function to infer the correlation, we directly test the causal
effect of technology adoption on staffing decisions.
Our work also contributes to the IT adoption literature on system performance. Lee et al.
(2013) documents the impact of IT investment on productivity. Dranove et al (2014) studies the
cost savings of health IT adoption. Niazkhani et al. (2009) reviews the studies about the adoption
of CPOE on clinical workflow. Most related to this study is a branch of the literature with an
emphasis on the impact of health IT adoption on the quality of care offered by medical providers
(Parente and McCullough, 2009, Queenan et al. 2011, Venkatesh et al. 2011, Aron et al. 2011,
Devaraj et al. 2013). Although these studies demonstrate a positive correlation between IT
adoption and quality, they do not consider the econometric issue that CPOE adoption is not an
exogenous event. Without correcting the endogeneity issue in adoption, the estimated results
may be biased. In this study, we introduce an IV approach to better infer the causal relation
between IT adoption and process quality (staffing decision).
6
3. Hypotheses Development
To develop testable hypotheses, we build a simple analytical model to analyze a nursing
home’s optimal staffing decision at different automation levels. We denote by the number of
staff in a nursing home and by the number of patients. Hence we can write the staff-to-patient
ratio as ≡ /. The staff-to-patient ratio is a major factor affecting the quality of care provided
by a nursing home (Konetzka et al. 2008, Lin 2014). An important factor that moderates the
quality of care provided by a nursing home with a given staff-to-patient ratio is the degree of IT-
enabled automation, which is represented in our model by a positive and continuous number, .
Given the staff-to-patient ratio and the automation level , the quality of care is = (, ).
We require to be increasing in both and . Note that because the optimal staffing ratio is
endogenous, the assumption on the monotonicity of does not suggest that an increase in
automation will necessarily lead to an increase in quality. For example, if an increase in
automation level leads to a decrease in staffing ratio , the overall effect on quality is
ambiguous. This monotonicity assumption thus states that ceteris paribus, nurses equipped with
automation technology can work more efficiently in a nursing home and spend more time with
patients, thus providing better care.
Although each state Medicaid program is different, the margin from an average
Medicaid patient is generally the lowest compared with Medicare or private patients. Because
prices for Medicare and Medicaid patients are heavily regulated by the government, a nursing
home’s quality of care is becoming an increasingly important factor in the competition for more
profitable patients. To focus on this unique industry characteristic, we model a nursing home’s
average revenue per patient as a function (, ) where is the quality of care offered by the
nursing home and ∈ [, ] is a parameter to model the heterogeneity among nursing homes.
We assume that (, ) is increasing in which implies that a nursing home with high-
quality care will have higher average revenue per patient. The rationale is that high-quality
nursing homes are more likely to attract more profitable patients (i.e., non-Medicaid patients)
than low-quality nursing homes and there are empirical evidences suggesting excess Medicaid
demand. For example, Ching et al. (2012) find that nearly 18% of the elderly who qualified for
Medicaid in the state of Wisconsin were rationed out and around 27% of Medicaid nursing
7
homes patients could not enter their first-choice nursing homes.9 To provide empirical evidence
of this model assumption, we use the 2006 enrollment information to illustrate the relation
between a nursing home’s staffing ratio and its quality mix. The top panel in Figure 1 shows the
distribution of quality mix across 100 quintile licensed nurse (LN) staffing ratios. As we can see,
the proportion of Medicare and privately paying consumers (the proxy for quality mix) increases
as the LN staffing ratio increases. Apparently, (, ) must satisfy lim ∂/ ∂ = 0 because
→
(, ) is necessarily capped by the revenue per patient from the most lucrative type of patients.
A stylized way to incorporate this requirement is to assume that (, ) is concave in , that is,
∂ / ∂ < 0. Finally, we denote the average wage paid to LNs by .
We assume a fixed occupancy rate which implies a fixed total number of patients for a
nursing home with a given number of beds. The rationale is that nursing homes can fill their
vacant beds with Medicaid patients who are typically in excess supply and the Certificate-ofNeed law restricts nursing home expansion. As the bottom panel in Figure 1 shows, average
occupancy rates remain stable over the years in spite of the increasing adoption rate of CPOE.10
This assumption simplifies our analysis and allows us to focus on the average revenue per patient
which is of particular importance to nursing homes. For ease of notation, we normalize the total
number of patients to 1.
To understand how an increase in automation affects the optimal staffing decision and
the implication for the quality of care provided, we treat the staffing decision as the decision
variable and treat the automation level as a parameter. Essentially, we are modeling how a
nursing home should optimally determine its quality level by choosing an appropriate staffing
level. Mathematically, we can write the surplus revenue optimization problem as follows:11
maxV() = (, ) − .
9
Because non-Medicaid patients are preferred to Medicaid, if there are not enough beds to serve all patients,
Medicare and private patients gain admission first and the remaining beds are allocated to Medicaid patients
(Nyman 1993).
10
In Section 7.1, we empirically test this assumption that the total number of patients does not vary much after the
CPOE adoption.
11
About 30% of nursing homes are nonprofit organizations or government owned. However, this does not mean that
these organizations do not intend to make a profit or care about their bottom lines. They operate in the same way as
for-profit nursing homes. Rather, a nonprofit organization does not distribute its surplus income to the organization’s
directors as profit or dividends. To avoid confusion, we use the term surplus revenue to denote the difference of
revenue and expense.
8
Note that because is not a decision variable in our model, the cost of investing in technology is
normalized to 0. The first-order condition for the nursing home’s optimization problem yields
∂(, )
= .
∂
Denote the solution to the optimization problem by ∗ , and the associated quality of care by ∗ .
Our analysis examines how an increase in automation level in a nursing home affects its
optimal staffing decision ∗ and the implication for the resulting quality ∗ . To this end, we
assume the following functional forms of (, ) and (, ):
(, ) = ,(, ) = 1 − & '() .
Clearly, (, ) and (, ) satisfy the following properties:∂/ ∂ > 0,∂/ ∂ > 0,∂/ ∂ >
0, ∂ / ∂ < 0, and lim ∂/ ∂ = 0. We impose the following technical assumption which
→
allows us later to interpret the parameter as the vertical position of a nursing home within its
competitive market:
& 0<<<+
.
,
From the first-order condition, we immediately have the following result.
Lemma: The optimal staffing level ∗ , the optimal quality level ∗ , and the resulting average
revenue per patient for a nursing home with vertical position are given below:
∗ =
1
, 1 , ln
, ∗ =
ln
,( ∗ , ) = 1 −
.
,
,
,
We introduced the parameter to model nursing home heterogeneity. Our next result
gives us a natural interpretation of as the vertical position of a nursing home within its market.
Proposition 1: The optimal staffing level ∗ , the optimal quality level ∗ , and the average
revenue per patient ( ∗ , ) are increasing in .
Proof: The claim follows immediately from the technical assumption and the following
expression,
, 2
−
ln
∂
.
=
∂
, ∗
9
□
To understand how an increase in automation affects the quality and the average revenue
per patient of a nursing home, we need to evaluate the signs of ∂ ∗ / ∂ and ∂( ∗ , )/ ∂
which is straightforward from Lemma 1 and is summarized in the next proposition.
Proposition 2: The optimal quality level ∗ and the average revenue per patient ( ∗ , ) are
increasing in the automation level .
Finally, to examine how an increase in automation affects the optimal staffing decision
of a nursing home, we need to evaluate the sign of ∂ ∗ / ∂ . The following proposition
summarizes the results.
Proposition 3: An increase in automation leads to an increase in a nursing home’s staffing level
01
01
if < / (2 , but it leads to a decrease in a nursing home’s staffing level if > / (2 .
Proof: The claim follows immediately from
, 1
−
ln
∂
.
=
∂
, ∗
□
The above result suggests that the effect of automation on the optimal staffing level is
more subtle than its effect on the optimal quality level. An increase in automation may lead to an
increase or decrease in the optimal staffing level depending on the vertical position of a nursing
home within its market. Intuitively, the use of automation makes health workers more productive,
so the marginal benefit in quality improvement from more staff increases. This complementarity
34 5
effect has its root in the property of the quality function (, ) (i.e., 36 32 > 0). Because quality
improvement eventually translates to a better quality mix and thus higher revenue, this
complementarity effect of automation encourages a nursing home to increase its staff. On the
other hand, although an increase in quality leads to better quality mix and thus increases revenue,
the marginal benefit of quality improvement for revenue diminishes as the quality level keeps
increasing because the average revenue per patient is concave in quality. The marginal cost of
staffing, however, is constant. Thus, for nursing homes that already have high quality (i.e., those
with higher values of , according to Proposition 1), an increase in automation may actually lead
to the substitution of technology for labor. The substitution effect has its root in the property of
the average price function (i.e.,
34 7
3 4
< 0 ). The co-existence of the two effects explains the
10
intuition behind Proposition 3: the complementarity effect dominates the substitution effect for
low-end nursing homes but is dominated by the substitution effect for high-end nursing homes.
We therefore propose the following hypotheses for empirical testing:
Hypothesis 1: An increase in automation leads to an increase in the quality level chosen by a
nursing home.
Hypothesis 2: An increase in automation leads to a decrease in the staff-to-patient ratio for a
nursing home with a high vertical position.
Hypothesis 3: An increase in automation leads to an increase in the staff-to-patient ratio for a
nursing home with a low vertical position.
4. Data and Empirical Methods
This study incorporates two primary data sources: the 2006-2012 OSCAR data and the
2005-2011 HIMSS data. The OSCAR data cover all Medicare- and Medicaid-certified nursing
homes operating throughout the United States. The database includes various characteristics, e.g.,
beds, payer types, patient health status, and staffing information. The HIMSS data provide
detailed information on different information technology applications adopted by many health
facilities, including nursing homes. We manually merged these two datasets using the name, zip
code and phone number of individual nursing homes. In total, we located 2,119 nursing homes
and constructed a seven-year, unbalanced panel with 12,313 observations.
In addition, we supplemented the OSCAR file with NHC data, which provides detailed
consumer complaint information for individual nursing homes, and Skilled Nursing Facility
(SNF) cost reports, which offer information on patient admissions. We also obtained the market
demographics from the Area Resource files. Table 1 reports the summary statistics of the key
variables
We assess the effect of CPOE adoption on nursing home process quality using the
following specification where the unit is a nursing home in a specific year:
89 : = ;< + ;> ?@9
,:'>
+ ; A9 : + ;B C9 : + ;D EFGF& ∗ 8&G: +;9 + ;: + H9
where 89 : represents process quality in nursing home I in state in year F; ?@9
,:'>
:
(1)
is a one-year
lagged binary variable that equals 1 if a nursing home adopts CPOE and 0 otherwise. Hence, the
coefficient ;> captures the effect of CPOE adoption on the dependent variable 89 : . In this twoway fixed effect model, A is a vector of nursing home characteristics, including the percentage of
11
Medicaid residents, patient health status, and total beds. C9
,:
is a vector of market characteristics
at the county level, including the intensity of competition as measured by the Herfindahl index,
the log of income per capita, and the log of the population. EFGF& ∗ 8&G: , a vector of variables,
represents state specific linear trends, which helps to control for the potential unobserved
trajectories at the state level. In addition, we control the nursing-home fixed effect (;9 ) for time-
invariant unobserved factors and the year effect (;: ) for yearly trends. ε is the error term. The
standard errors are clustered by nursing home.
5.2 Specification for Staffing
We are most interested in how CPOE adoption affects staffing for nursing homes with
different vertical positions. We use the following specification to estimate the different effects of
CPOE adoption on nursing home staffing:
E9 : = K< + K> ?@9
,:'>
+ K ?@9
,:'>
∗ LMIFIMN9
<
+ KB A9 : + KD C9
+KO EFGF& ∗ 8&G: +K9 + K: + H9
:
This specification is similar to specification (1). The dependent variable E9
:
(2)
:
represents the
staffing level in nursing home I in state in year F. We add in an interaction term by interacting
the initial distance of each nursing home, Position, with its CPOE adoption, IT. The coefficient
K> captures the effect of CPOE adoption on staffing for a nursing home whose initial licensed
nurse staffing level is binding at the state minimum staffing ratio. If K> > 0, this suggests that the
adoption of CPOE increases staffing in nursing homes at the lowest end in a local market. The
coefficient K captures the differential effect of CPOE adoption. If K < 0, this indicates that a
nursing home with higher vertical quality is associated with less growth (or greater reduction) in
staffing due to the CPOE adoption. Hence, K> + K ∗ PMIFIMN captures the CPOE effect on
staffing decisions in nursing homes with better vertical quality.
5.3 Identification and Instrumental Variable
The econometric challenge in estimating the coefficients of the specifications above is that
the adoption of CPOE may be correlated with other (unobserved) decisions related to staffing or
process quality. For example, the managerial incentives in a nursing home could simultaneously
affect decisions about both the adoption of CPOE and staffing. There might be a reverse
causality issue, too. For example, nursing homes with high quality may be more likely to adopt
12
CPOE to maintain their high quality. Therefore, we have to consider the endogeneity for
unbiased estimations.
To establish a causal link between CPOE adoption and the staffing (quality) decision in
nursing homes, we introduce an instrumental variable approach in this study. Miller and Tucker
(2009) show the network effect that greater adoption by other hospitals should lead to greater
network benefits for health IT. The network benefits may lead to learning effects (Karshenas and
Stoneman, 1993) and also reduce the costs of transferring information across the network. As a
result, hospitals are responsive to past adoption by other local hospitals. Following Miller and
Tucker (2009), we construct an instrumental variable, hospital_CPOE, describing the yearly
hospital CPOE adoption rates in the local market where we define a county as a market. We
argue this is a valid instrumental variable because in the healthcare value chain, hospitals are
upstream while nursing homes are downstream. All the Medicare patients and some of the other
patients are transferred from hospitals to nursing homes for post-acute care. Nursing homes thus
share the same network effects with hospitals in the local market, especially the ease of
transferring information. However, the adoption behavior of all the hospitals in the local market
should not directly affect the quality and staffing decisions of individual nursing homes.
6. Empirical Results
In this section, we report our empirical results regarding the impact of CPOE adoption on
process quality and staffing in nursing homes. We first show that the number of resident
complaints about quality delivery is negatively correlated with the adoption of CPOE, thus
supporting Hypothesis 1. Next, we show that the effect of CPOE adoption on a nursing home
depends on the nursing home’s vertical position. Consistent with Hypotheses 2 and 3, we find
that after the implementation of CPOE, high-end nursing homes reduce staffing while low-end
nursing homes increase staffing. This section ends with a robustness check using an alternative
instrumental variable.
6.1 Process Quality
Table 2 reports the impact of the adoption of CPOE on consumer satisfaction. This table
covers results using different measures of resident complaints as the dependent variable. The first
three columns use the total complaint counts as the dependent variable. Column (1) reports the
13
OLS results. Column (2) shows the results in the first-stage estimation, followed by the 2SLS
results in Column (3). The OLS coefficient of CPOE in Column (1) is -0.05 and insignificant.
After implementing the instrumental variable, the coefficient becomes -0.51 at the one percent
significance level. This means that adopting CPOE can result in 0.51 fewer complaints per year
on average, which is equivalent to a 36.9% reduction in annual resident complaints.
The difference in the coefficients between OLS and 2SLS is very large though the two
coefficients share the same sign. There could be many reasons due to the endogeneity in CPOE
adoption. One candidate explanation is that the majority of the adopters initially have low
resident satisfaction or a large number of complaints, which could cancel off the effect of CPOE
and bias the OLS estimates.
We are also concerned that the results may be driven by those nursing homes with many
complaints. To relieve this concern, we repeat the same estimation model, but use an alternative
complaint count measure for which we drop the outliers and trim the raw complaint counts at the
99th percentile. Column (4) shows the results using the trimmed dependent variable: the
coefficient of CPOE is 0.397 with p < 0.05, smaller than that in our main results.
Overall, Tables 2 shows that the adoption of CPOE strongly increases resident satisfaction.
The improvement occurs across all domains and across different severity levels. In particular, the
implementation of CPOE significantly reduces complaints about administrative issues and
prevents residents from being harmed severely. These findings are consistent with our
Hypothesis 1.
6.2 Staffing
To test Hypotheses 2 and 3, we analyze the changes in staffing across nursing homes.
We differentiate nursing homes by their initial vertical positions. Table 3 reports the
heterogeneous effects of CPOE adoption on staffing using specification (2), with an emphasis on
LNs. The dependent variables are nurse hours per resident day (HPRD) for LNs and RNs
respectively. We report the results using OLS and 2SLS following specification (2).
The coefficient of CPOE in Column 2 is 0.282 at one percent significance level,
suggesting that nursing homes with the lowest staffing levels in their markets hire more LNs
after the adoption of automated technology. The coefficient of the interaction term,
CPOE*Position, is negative and significant. The coefficient suggests that an increase by one
14
standard deviation in initial staffing distance (Position) reduces the staffing increment by 0.170
HPRD compared with the base outcome. When the Position is above a certain threshold, the
nursing home may use fewer LNs.
In Column 3, we replace the continuous vertical position measure with a binary variable
as a robustness check. The positive coefficient of CPOE suggests that low-end nursing homes
increase LN staffing by 0.145 HPRD or 7.6% from the mean. By contrast, high-end nursing
homes actually use 0.255 HPRD fewer LN staff than low-end nursing homes. The joint F test
shows that CPOE adoption reduces the use of LNs by 0.110 HPRD in high-end nursing homes, a
5.8% reduction from the mean.
To show that our results are not sensitive to the choice of minimum staffing ratio, we
also use RN minimum staffing ratios as alternative cutoffs for the calculation of vertical position.
Although only a few states impose separate mandates for RNs, Harrington (2010) estimated
different state standards to a uniform format—RN hours per resident day for a 100-bed nursing
home. Columns (4) and (5) report the heterogeneous effects using the estimated RN minimum
staffing ratios. The results remain robust both quantitatively and qualitatively.
In general, the results in Table 3 support Hypotheses 2 and 3, that low-end nursing
homes increase their staff while high-end ones decrease their staff after the implementation of
CPOE. As a strategic provider, low-end nursing homes adopting new technologies invest more in
nurses to leverage their competitiveness in the entire market, while high-end nursing homes have
incentives to cut staff in order to contain costs.
8. Conclusion
We study the effect of IT-enabled automation on the staffing decisions and process quality
of healthcare providers using a unique dataset covering 2,119 surveyed nursing homes in the U.S.
over a seven-year period. We find that the adoption of IT-enabled automation technology
decreases staffing in the high-end nursing homes while increases staffing in the low-end. We
also find that the adoption of advanced information technology reduces resident complaints
about process quality by 36.8% on average, which results in the change of resident composition
by a 14.7% decrease in the admissions of Medicaid patients, the least profitable type, regardless
of the nursing home’s vertical position. Further, we rule out an alternative argument that the
15
efficiency is mainly gained through digitization. All these results are consistent with the
predictions of a staffing model that incorporates technology adoption and vertical differentiation.
Our findings suggest that the relation between staffing and automation adoption is mixed.
It crucially depends on the relative strength of the two opposing effects: complementarity and
substitution, across different types of nursing homes. Our theoretical model indicates that the
complementarity effect dominates the substitution effect in the low-end nursing homes. To
attract lucrative patients, these low-end nursing homes have great incentives to hire more nurses
to increase their competitiveness in the local marketplace after they adopt the automation
technology. On the contrary, the substitution effect dominates the complementarity effect in the
high-end nursing homes. Since automation technology significantly increases the utilization of
nurse time and the marginal benefit of providing additional quality is relatively low, these highend nursing homes are likely to reduce their staffing level for the sake of cost containment. To
the best of our knowledge, this is one of the few papers in the information system literature to
study the impact of automation technology adoption on a firm's strategic staffing decisions.
One potential limitation of this study is the limitation of the HIMSS data which cover a
small set of nursing homes. Although nursing homes in the HIMSS sample share similar
occupancy rates and patient profiles with the entire nursing home population, surveying a
broader set of nursing homes may be helpful for other IT-related nursing home research.
Our findings have important implications. For individual nurses, automation technology
does not necessarily result in nurse job loss. Nurses can anticipate their prospective employment
status by recognizing the vertical position of a nursing home where they are working or will
work. For a nursing home, the effect of technology adoption on staffing decision largely depends
on its vertical position in the marketplace. Managers can have a clear strategy in the staff
planning when IT adoption becomes the new trend and follow either a revenue expansion
strategy or a cost reduction strategy depending on the vertical position of the nursing home. For
policy makers, our results show that adoption of information technology improves quality in the
nursing home industry, so the government should provide subsidies to nursing homes for
adoption.
16
References
Acemoglu, D. 2002. "Technical Change, Inequality, and the Labor Market." Journal of
Economic Literature, 40(1), 7-72.
Acemoglu, D. and A. Finkelstein. 2008. "Input and Technology Choices in Regulated Industries:
Evidence from the Health Care Sector." Journal of Political Economy, 116(5), 837-80.
Agarwal, R., G. D. Gao, C. DesRoches and A. K. Jha. 2010. "The Digital Transformation of
Healthcare: Current Status and the Road Ahead." Information Systems Research, 21(4), 796-809.
Aksin, Z., M. Armony and V. Mehrotra. 2007. "The Modern Call Center: A Multi-Disciplinary
Perspective on Operations Management Researc h." Production and Operations Management,
16(6),665-688.
Anderson, C. L. and R. Agarwal. 2011. "The Digitization of Healthcare: Boundary Risks,
Emotion, and Consumer Willingness to Disclose Personal Health Information." Information
Systems Research, 22(3), 469-90.
Angst, C. M. and R. Agarwal. 2009. "Adoption of Electronic Health Records in the Presence of
Privacy Concerns: The Elaboration Likelihood Model and Individual Persuasion." MIS Quarterly,
33(2), 339-70.
Angst, C. M., S. Devaraj, C. C. Queenan and B. Greenwood. 2011. "Performance Effects Related
to the Sequence of Integration of Healthcare Technologies." Production and Operations
Management, 20(3), 319-33.
Aron, R., S. Dutta, R. Janakiraman and P. A. Pathak. 2011. "The Impact of Automation of
Systems on Medical Errors: Evidence from Field Research." Information Systems Research,
22(3), 429-46.
Autor, D. H., L. F. Katz and A. B. Krueger. 1998. "Computing Inequality: Have Computers
Changed the Labor Market?" Quarterly Journal of Economics, 113(4), 1169-213.
Berman, E., J. Bound and Z. Griliches. 1994. "Changes in the Demand for Skilled Labor within
United-States Manufacturing - Evidence from the Annual Survey of Manufactures." Quarterly
Journal of Economics, 109(2), 367-97.
Brickley, J., S.F. Lu and G. Wedig, "Malpractice Laws and Incentives to Shield Assets: Evidence
from Nursing Homes", working paper
Castle, N. G. 2008. "Nursing Home Caregiver Staffing Levels and Quality of Care - a Literature
Review." Journal of Applied Gerontology, 27(4), 375-405.
Ching, A., F. Hayashi, and H. Wang. 2014. "Quantifying the Impacts of Limited Supply: the
Case of Nursing Homes." International Economic Review, forthcoming.
17
Chwelos, P., R. Ramirez, K.L. Kraemer and N. P. Melville. 2010. "Does Technological Progress
Alter the Nature of Information Technology as a Production Input? New Evidence and New
Results." Information Systems Research, 21(2), 392-408.
Credit Suisse Equity Research. 2001. "Long-Term Care Industry."
Davidson, E. J. and W. G. Chismar. 2007. "The Interaction of Institutionally Triggered and
Technology-Triggered Social Structure Change: An Investigation of Computerized Physician
Order Entry." MIS Quarterly, 31(4), 739-58.
Devaraj, S., T. T. Ow and R. Kohli. 2013. "Examining the Impact of Information Technology
and Patient Flow on Healthcare Performance: A Theory of Swift and Even Flow (Tsef)
Perspective." Journal of Operations Management, 31(4), 181-92.
Dewan, S. and C. K. Min. 1997. "The Substitution of Information Technology for Other Factors
of Production: A Firm Level Analysis." Management Science, 43(12), 1660-75.
Dranove, D., C. Forman, A. Goldfarb and S. Greenstein. 2014. "The Trillion Dollar Conundrum:
Complementarities and Health Information Technology." American Economic Journal:
Economic Policy, 6(4), 239-270.
Eisenberg, F. and A. S. Barbell. 2002. "Computerized Physician Order Entry: Eight Steps to
Optimize Physician Workflow." Journal of Healthcare Information Management, 16(1), 16-18.
Felan, J. T., T. D. Fry and P. R. Philipoom. 1993. "Labor Flexibility and Staffing Levels in a
Dual-Resource Constrained Job-Shop." International Journal of Production Research, 31(10),
2487-506.
Furukawa, M, T.S. Raghu and B. Shao. 2010. "Information Technology, Skilled Labor, and
Factor Substitution: Evidence from the Healthcare Industry." Workshop on Information Systems
and Economics.
Galanter, W. L., R. J. Didomenico and A. Polikaitis. 2005. "A Trial of Automated Decision
Support Alerts for Contraindicated Medications Using Computerized Physician Order Entry."
Journal of the American Medical Informatics Association, 12(3), 269-74.
Garfinkel, R., R. Gopal and S. Thompson. 2007. "Releasing Individually Identifiable Microdata
with Privacy Protection against Stochastic Threat: An Application to Health Information."
Information Systems Research, 18(1), 23-41.
Goh, J. M., G. D. Gao and R. Agarwal. 2011. "Evolving Work Routines: Adaptive Routinization
of Information Technology in Healthcare." Information Systems Research, 22(3), 565-85.
Griliches, Z. 1969. "Capital-Skill Complementarity." Review of Economics and Statistics, 51(4),
465-68.
18
Grabowski, D.C., J. Gruber and J.J. Angelelli. 2008. "Nursing Home Quality as a Common
Good." Review of Economics and Statistics, 90(4), 754-764.
Kane, G. C. and G. Labianca. 2011. "Is Avoidance in Health-Care Groups: A Multilevel
Investigation." Information Systems Research, 22(3), 504-22.
Karshenas, M. and P. L. Stoneman. 1993. "Rank, Stock, Order, and Epidemic Effects in the
Diffusion of New Process Technologies - an Empirical-Model." Rand Journal of Economics,
24(4), 503-28.
Konetzka, R. T., S. C. Stearns and J. Park. 2008. "The Staffing-Outcomes Relationship in
Nursing Homes." Health Services Research, 43(3), 1025-42.
Krusell, P., L. E. Ohanian, J. V. Rios-Rull and G. L. Violante. 2000. "Capital-Skill
Complementarity and Inequality: A Macroeconomic Analysis." Econometrica, 68(5), 1029-53.
Lahiri, A. and A. Seidmann. 2012. "Information Hang-Overs in Healthcare Service Systems."
Manufacturing & Service Operations Management, 14(4), 634-53.
Lee, J., J. S. McCullough and R. J. Town. 2013. "The Impact of Health Information Technology
on Hospital Productivity." Rand Journal of Economics, 44(3), 545-68.
Li, Y., C. Harrington, W. D. Spector and D. B. Mukamel. 2010. "State Regulatory Enforcement
and Nursing Home Termination from the Medicare and Medicaid Programs." Health Services
Research, 45(6), 1796-814.
Lin, W. T. and B. B. M. Shao. 2006. "The Business Value of Information Technology and Inputs
Substitution: The Productivity Paradox Revisited." Decision Support Systems, 42(2), 493-507.
Lu, S. F. and L.X. Lu. 2014. "Does Mandatory Overtime Law Improve Quality? Staffing and
Operational Flexibility in Nursing Homes " Working Paper.
Menon, N. M., B. Lee and L. Eldenburg. 2000. "Productivity of Information Systems in the
Healthcare Industry." Information Systems Research, 11(1), 83-92.
Miller, A. R. and C. E. Tucker. 2009. "Privacy Protection and Technology Diffusion: The Case
of Electronic Medical Records." Management Science, 55(7), 1077-93.
Mishra, A. N., C. Anderson, C. M. Angst and R. Agarwal. 2012. "Electronic Health Records
Assimilation and Physician Identity Evolution: An Identity Theory Perspective." Information
Systems Research, 23(3), 738-60.
Niazkhani, Z., H. Pirnejad, M. Berg and J. Aarts. 2009. " The Impact of Computerized Provider
Order Entry Systems on Inpatient Clinical Workflow: A Literature Review." Journal of the
American Medical Informatics Association, 16(4), 539-549.
19
Nyman. 1993. "Testing for Excess Demand in Nursing Home Care Markets." Medical Care,
31(8), 680-693.
Northon, E. C. 2000. "Long-term Care." Handbook of Health Economics. 1, 95-994.
Oborn, E., M. Barrett and E. Davidson. 2011. "Unity in Diversity: Electronic Patient Record Use
in Multidisciplinary Practice." Information Systems Research, 22(3), 547-64.
Ozdemir, Z., J. Barron and S. Bandyopadhyay. 2011. "An Analysis of the Adoption of Digital
Health Records under Switching Costs." Information Systems Research, 22(3), 491-503.
Parente, S. T. and J. S. McCullough. 2009. "Health Information Technology and Patient Safety:
Evidence from Panel Data." Health Affairs, 28(2), 357-60.
Queenan, C. C., C. M. Angst and S. Devaraj. 2011. "Doctors' Orders--If They're Electronic, Do
They Improve Patient Satisfaction? A Complements/Substitutes Perspective." Journal of
Operations Management, 29(7-8), 639-49.
Venkatesh, V., X. J. Zhang and T. A. Sykes. 2011. ""Doctors Do Too Little Technology": A
Longitudinal Field Study of an Electronic Healthcare System Implementation." Information
Systems Research, 22(3), 523-46.
Worth, T. 2008. "Nurses Have Little Time for Patients." American Journal of Nursing, 108(9),
20-21.
Zlabek, J. A., J. W. Wickus and M. A. Mathiason. 2011. "Early Cost and Safety Benefits of an
Inpatient Electronic Health Record." Journal of the American Medical Informatics Association,
18(2), 169-72.
20
.2
.3
Quality Mix (%)
.4 .5 .6 .7
.8
Figure 1: Quality Mix and Occupancy Rates
20
30
40
50
60
70
Percentiles of Staff-to-Patient Ratio
80
90
100
Occupancy(%)
.5 .6 .7 .8 .9
1
10
2006
2007
2008
2009
Year
National Average
21
2010
Adopter Average
2011
2012
Table 1: Summary Statistics
Variable
Technology Adoption
CPOE
Measures of Process Quality
Total Complaints
Complaints about
Administration
Complaints about Quality of
Care
Complaints about Quality of
Life
Complaints about Minor
Issues
Complaints about Major
Issues
Staffing Measures
LN HPRD
RN HPRD
Vertical Position
Distance to MSR
Control Variables
Percentage of Medicaid
Beds
ADL Index
HHI Competition
Log Income
Log Elderly Population
Mean
SD
Definition
0.20
0.40
1 if CPOE was adopted last year
1.38
3.11
total counts of consumer complaints
0.64
1.58
complaints in the domain of administration
0.44
1.07
complaints in the domain of quality of care
0.30
0.99
1.16
2.76
complaints in the domain of quality of life
complaints for minor harm and potential
harm
0.21
0.78
complaints for actual harm and jeopardy
1.91
1.00
0.97
0.74
licensed nurse hours per resident day
registered nurse hours per resident day
1.42
0.99
initial distance to minimum staffing
requirement
0.55
100.7
8.21
0.03
10.46
9.76
0.27
79.3
3.68
0.03
0.57
1.67
percentage of Medicaid patients
total beds (weighted by 100 in regressions)
ADL index describing patient health status
competition measure at the county level
log of income per capita at the county level
log of elderly population at the county level
22
Table 2: Effects of Adoption of CPOE on Process Quality
Dependent Variable:
Resident Complaints
on Process Quality
CPOE
Percentage of Medicaid
Beds
ADL Index
HHI Competition
Log Income
Log Elderly Population
IV: Hospital_CPOE
Baseline
Trimming
Total Complaints
Outliers
First
OLS
Stage
2SLS
2SLS
(1)
(2)
(3)
(4)
-0.046
-0.508***
-0.361**
(0.080)
(0.196)
(0.174)
-0.149
0.011
-0.134
-0.195
(0.321)
(0.026)
(0.320)
(0.278)
0.456**
-0.028
0.452**
0.494**
(0.229)
(0.029)
(0.224)
(0.207)
0.025
-0.001
0.024
0.023
(0.018)
(0.002)
(0.018)
(0.015)
3.825
0.013
3.929
4.300
(6.021)
(0.623)
(6.041)
(5.257)
-0.428
-0.046
-0.497
-0.242
(1.207)
(0.084)
(1.203)
(0.882)
-0.968
0.157
-0.918
-0.811
(0.985)
(0.106)
(0.979)
(0.825)
0.552***
(0.022)
Y
Y
Y
Y
Y
Y
Y
Y
Nursing Home Dummies
Year Dummies
Individual State Linear
Trends
Y
Y
Y
Y
Weak Identification Test
Kleibergen-Paap rk Wald F statistic: 622.17***
Observations
12313
12313
12250
12250
Within R-squared
0.017
0.272
0.015
0.015
Number of provider
2119
2119
2056
2056
Robust standard errors in parentheses clustered by nursing
home
*** p<0.01, ** p<0.05, *
p<0.1
23
Table 3: The CPOE Adoption Effects on Staffing Decisions
Dependent Variable:
Hours per Resident Day
CPOE
CPOE * Position
CPOE * High End
F test: CPOE+CPOE* High End
Nursing Home Dummies
Year Dummies
State Linear Trends
Time Varying Controls
Observations
Within R-squared
Number of provider
Licensed Nurses
Registered Nurses
Minimum LNs
Minimum RNs
OLS
2SLS
2SLS
2SLS
2SLS
(1)
(2)
(3)
(4)
(5)
0.106*** 0.282*** 0.145*** 0.154*** 0.073**
(0.036)
(0.062)
(0.046)
(0.040)
(0.029)
-0.065** -0.172***
-0.145***
(0.029)
(0.042)
(0.044)
-0.255***
-0.109**
(0.071)
(0.047)
-0.110**
-0.036*
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
12,313
12,250
12,250
12,250
12,250
0.046
0.040
0.041
0.057
0.058
2,119
2,056
2,056
2,056
2,056
Robust standard errors in parentheses clustered by nursing home
*** p<0.01, ** p<0.05, * p<0.1
24