ASYMMETRIC INFORMATION IN THE MARKET FOR MEDICINE:

The Relationship Between Patient Satisfaction and Hospital Quality

Cristobal Young,

Department of Sociology, Stanford University

Xinxiang Chen,

National Strategic Planning & Analysis Research Center, Mississippi State University

Word Count: 12,152

September 17, 2011

Working Draft. Please do not cite or circulate without permission.

ABSTRACT:

The paper analyzes information problems in the context of hospital medical care. Using a large

sample of US hospitals, we find that patients have limited ability to gauge the quality of their

treatment. Raising the quality of medical care by 10 percent leads to only a 1.1 percent increase

in patient satisfaction. In contrast, patients are more sensitive to “room and board” aspects of

care: having a quiet room matters just as much for patient satisfaction as technical medical

quality, and positive interaction with nurses has an effect size many times larger than medical

quality. If hospitals compete on the basis of patient satisfaction, they will face an incentive to cut

back on medical quality and invest in the more fleeting and superficial (but highly visible)

aspects of care. This suggests caution in the push to reform medicine around market principles.

2

Introduction

Marketization is a broad movement in the United States pushing for a greater role for

market forces and incentives in the social organization of many areas of life that have long been

relatively non-market in nature. Areas such as public schools (Ladd 2002; Levin and Belfield

2003), social insurance programs such as Social Security and Unemployment Insurance

(Feldstein 2005), and health care (Herzlinger 1997; 2004) have all seen growing pressures to

reform along market principles. Advocates of such reforms insist that competition leads to a race

to the top, bringing better outcomes and lower costs. Advocates also advance an empowerment

ideology emphasizing the expansion of choice and agency under market-based structures.

However, where emergent markets suffer problems of quality uncertainty and

asymmetric information, marketization may bring significant downside risks to “consumers”.

Markets will deliver what consumers can observe and financially reward, which may not be the

same as what consumers want or need.

In health care, market proponents insist that better outcomes would emerge if health care

was more driven by patients‟ own dollars. If patients came to physicians with “cash in hand,”

they would demand better and more cost-effective treatment. This is the perspective embraced by

the “consumer-driven health care” movement in the United States. Critics argue that this

promises a caveat emptor, “buyer beware” market where professional ethics and fiduciary

responsibility fade and doctors and hospitals come to focus on selling whatever patients are

willing to pay for (Berenson and Cassel 2009). Central to this is the axis of “patient / consumer”

and “physician / seller”. Should those receiving medical treatment be thought of as “patients,” or

“consumers”? Should the providers in a medical market be thought of as “doctors,” or “sellers”?

The fundamental question is the extent to which these players should be exposed to market

3

incentives. Should health care reform focus on establishing greater sensitivity to market forces?

Are there downside risks to marketization? Should we, instead, seek to insulate patients and

doctors from market incentives?

This paper focuses on information problems in a health care system under pressure to

become more like a classical (text-book) free market. We emphasize that patients (consumers)

face a problem of partial information. They have limited ability to observe the technical quality

of their medical care, but they can observe the hospitality / room and board aspects of their care

quite well. If hospitals are to compete for patients and “customer loyalty”, there are likely to be

different returns to investing in medical quality versus investing in patient hospitality. Hospitals

that learn how to provide a warm emotional experience for patients may financially out-perform

those that primarily excel at medical quality. Marketization, in this context, may introduce

perverse incentives for hospitals that distract from their core mission of medical excellence.

Empirically, this paper draws on a large sample of American hospitals, with information

about both the technical quality of medical care and the hospitality / room and board aspects of

care. We test the degree to which patients can identify (are more satisfied in) hospitals with

better quality medical care.

The Partial Information Problem.

Hospitals face the challenge of balancing two general tasks: providing technical medical

treatment, and hospitality or “room and board” care while the patient lives in the hospital. At a

professional level, these tasks often run in an opposite direction. Some types of medical

treatment provide immediate relief from suffering. But more often, medical intervention is

painful and unpleasant, sacrificing short-term well-being for long-term gains in health status,

4

physical functioning and life expectancy. This is why sick and injured patients allow themselves

to be cut open, radiated, and exposed to toxins such as chemotherapy and other cocktails of

potent medication.

The other aspect of hospital treatment involves less-technical, more mundane care for the

patient while they live in the hospital: the “room and board” aspect of care. Patients must be fed

(is the food warm, tasteful?), they must sleep (is the room quiet or overrun?), they must cope

with their immediate pain, anxiety, fears, and frustrations (are the nurses and staff kind and

compassionate, generous with pain medication, quick to respond to problems?).

Patients generally lack expertise in medicine. When they become ill or injured, they rely

on expert medical judgment to know what is wrong with them, what treatment they need, and

what doctor would best provide that treatment (Arrow 1963; Parsons 1951; Akerlof 1970). This

diverges from the prototypical market transaction where customers are “king,” where consumers

decide what they want and then buy from the seller that offers the best product at the lowest

price. In the market for medicine, patients are unsure what they need, unsure which product is

best, and thus uneasy about buying the cheapest medical service. In practice, patients rely on

“sellers” of medical services to tell them what they need and what they should buy.

In contrast, the quality of “room and board” care in hospitals is exceedingly (and

sometimes painfully) visible to patients. They know when the food is cold and tasteless, when

their room is loud and overcrowded, when the nurses and staff are indifferent or too busy to care

about their pains and problems. For these hospitality aspects of hospital care, there is no

objective metric beyond the patient‟s own assessment of their experience. When hospitals fail on

these fronts, there is no expert who can explain to patients that they did, in fact, receive excellent

“room and board” care.

5

This is a more general market issue where some aspects of a good or service are hard to

observe: a good or service is expected to “embody several ideal standards” that vary in terms of

their observability (Goffman 1959:44). One might think of this as partial asymmetry of

information, or partial uncertainty about quality.

Patients may think of the world as having “good” hospitals and “bad” hospitals, and then

use whatever information they can observe to categorize their hospital. Unpleasant nurses and

busy, noisy rooms, then, become evidence that one is in a “bad” hospital. Lynn (2006)

emphasizes how, under conditions of uncertainty, “auxiliary characteristics become proxies for

quality”. Room and board / hospitality care becomes a proxy for the (unobservable) medical

quality that will most impact a patient‟s life. This is a process of inferring from the known to the

unknown, or from the observed to the unobserved.

This type of process has been studied in different settings. In social status research, Lynn,

Podolny and Tao (2009) analyze how quality uncertainty can lead to the “decoupling” between

an individual‟s merit and their social status. Podolny (1993) finds that, in the investment banking

market, a banker‟s status or reputation becomes a more valuable asset as market uncertainty

increases. Lynn (2006) finds that in academic disciplines where research quality is harder to

evaluate, journal article citations are more influenced by extraneous factors such as the prestige

of an author‟s university or the length of the article. A large body of research shows there is a

“halo effect” of beauty, in which physically attractive people are regarded as more intelligent,

competent, and trustworthy. In labor market hiring decisions, Dovidio and Gaertner (2000) find

that when an applicant‟s qualifications are ambiguous, race plays a much stronger role in

evaluating the applicant‟s merit (with white applicants much more likely to be hired). In used car

markets, buyers often consider the cleanliness of a car as evidence of its mechanical condition. In

6

all of these cases, individuals are using technically unrelated or extraneous information to “fill

in” important gaps in their knowledge in order to make decisions.

Even when quality is observable, extraneous information can still be distracting, and lead

actors into poor decisions. For example, the introduction of blind auditions among elite

symphony orchestras substantially increased the likelihood that female musicians would advance

and be hired in competitive auditions (and reduced the probability that male musicians would

advance) (Goldin and Rouse 2000). The knowledge of a musician‟s gender weakened

recruitment committees‟ ability to accurately identify quality. This is a case where excessive

weight is placed on distracting information.

Hospital care is a ripe area where one aspect of service (hospitality) can serve as a signal

for another (medical quality). There is a potentially serious problem of encouraging

marketization when only some aspects of quality are observable – particularly when the most

important aspect of quality is hardest to observe. If hospitals compete on the basis of patient

satisfaction, they may face an incentive to cut back on (hard to observe) medical quality and

invest more in superficial (ie, observable but less important) aspects of care. This is particularly

so in an environment where there is much concern about cost inflation and the need to limit

health care expenditures. The combination of marketization of health care, partial information

problems, and cost control may mean that technical medical quality may be crowded out by

market forces that reward hospitality more than medical quality.

7

Market-Driven Health Care

A prominent movement in the United States advocates market driven health care

(Herzlinger 1997), also frequently known as “consumer driven” health care (Herzlinger 2004;

Scandlen 2004).1 Proponents argue that there are important benefits to shifting costs to patients

that will “activate” patients into consumers and “drive a new quality paradigm” (Retchin

2007:173).

The central concern is that Americans are over-insured, and do not spend enough of their

own money out of pocket. By letting insurance companies pay for their treatment, patients

relinquish their role as discriminating consumer – a missed opportunity to control costs and

improve quality in a more competitive and responsive market for medicine.

Over-insurance has “prevented [patients] from developing the habit of discernment

necessary to make smart choices” (Naifeh and Smith 2004:463). When patients are spending

their own money, and thus making trade-offs between cost, quantity, and quality of care, they

will become better managers of their medical problems. Rather than simply following doctors‟

advice and letting insurance pay the bill, patients who are spending their own money will think

more carefully about the value of their treatment, and like many fickle and hard-to-please

consumers, will demand more for their money: better treatment, more careful diagnosis, smarter

and more efficient care. With limited insurance and their own savings, patients become

“consumers with cash-in-hand, demanding to know for themselves who is the best urologist in

town, what are my treatment alternatives, … where is the nurse when I need one, how do I get

the most value for the money I‟m spending” (Scandlen 2004:1117). As leading advocate Regina

1 The consumer driven health care movements needs to be distinguished from general groups advocating patients‟

rights, voice, and empowerment (Cleary 2003).

8

Herzlinger insists, “health care will not improve until consumers drive it” (Herzlinger

2004:XXIII).

The policy solution is high-deductible, or “catastrophic coverage,” insurance, in which

patients pay the full cost of their treatment up to some deductible (ranging from $2,400 up to

$12,000). These limited-insurance health plans barely existed in the mid-1990s, but are now

offered by virtually every insurance company, and show rapidly growing enrollments –

particularly among large employers (AHIP 2010).2 The high-deductible market has expanded

from under half a million persons covered in 2004, to over 10 million people by 2010 (AHIP

2010).3

Source: American Health Insurance Providers (AHIP 2010).

2 The Bush administration gave a strong boost to these plans with the Medicare Modernization Act of 2003, which

established tax-sheltered Health Savings Accounts for those with a high-deductible insurance plan. Similar to an

Individual Retirement Account (IRA), individuals can “accumulate tax-free savings to pay for medical expenses”

(GAO 2008:1). In short, the federal government subsidizes the price of medical treatment for those on a high-

deductible plan. The subsidy is equal to the individual‟s marginal tax rate, multiplied by the portion of medical

expenses paid from the health savings account. In other words, the subsidy is highest for people in the highest tax

brackets who save the most. 3 Currently, the plans require families to pay at least $2,400 a year in out-of-pocket medical expenses before

insurance coverage begins, and deductibles can go as high as $12,000 per year. The average deductible in 2010 was

$4,000 (AHIP 2010).

0

2

4

6

8

10

12

2004 2005 2006 2007 2008 2009 2010

Persons covered by High-Deductible Insurance Plans, 2004-10

Millions

9

How effectively can patients-as-consumers “drive a new quality paradigm” (Retchin

2007:173)? What are the downside risks of a limited-insurance, more market-based health care

system? These questions hinge on the ability of patients to identify and reward better quality

medical care.

A more incentivized market will not necessarily give patients get what they want and

need, but rather what they can observe, monitor, and financially reward. Can patients tell when

medical management is going down the wrong path, and when they should start to demand more

for their money? The ability to observe the quality of medical treatment is central to whether

“activated,” price-sensitive patients will produce improvements in health care.

Problems in the Technical Quality of Medical Care

Patients often take the technical quality of their medical care for granted, and usually (at

least in the absence of disconcerting signals) “assume technical competence at both professional

and organization levels” (Elwyn et al 2007:1021). However, patients have good reason to worry

about the technical quality of their medical care. Mistakes in medicine are frequent and carry

severe risks. Rates of medical injury are high. Doctors often fail to apply up-to-date or

recommended standards of care. These factors make hospitals dangerous places to reside.

Research indicates that doctors often fail to apply the full recommended treatment for

many classes of illness (McGlynn, et al 2003; RAND 2006). After reviewing 6,700 medical files

from across the country, RAND estimated that little more than half (55%) of patients receive the

full recommended treatment for their ailments. Quality varies by illness; some 65% of people

with hypertension received the recommended care, though only 40% of pneumonia patients

10

received proper treatment. As the RAND study gently noted, “patients should not assume that

their physicians will remember all that needs to be done” (RAND 2006:5).

These shortcomings in treatment are usually not catastrophic. However, large-scale

studies of hospital patients find that about 3 to 4 percent of hospitalizations lead to an “adverse

event”: a disabling injury caused by medical treatment (Leape et al 1991; de Vries, et al 2008).4

In the Harvard Medical Practice Study of 30,000 patient medical records (Brennan et al 1991;

Leape et al 1991), the most common adverse events were drug complications, wound infections,

and technical surgical complications. [More dramatic errors include “retained instruments” left

inside of patients after surgery, operations on the wrong limb or patient, and so on (Wachter and

Shojania 2004).] Most medical injuries caused only short-term disability (lasting less than six

months), although 3 percent caused permanent disability, and 14 percent resulted in death.5

Roughly 60 percent of the injuries were clearly related to medical errors, and 28 percent met a

legal criterion of negligence (in the judgment of two independent physician-reviewers).

Extrapolating these results to the 35 million hospital admissions in the US in 2004 suggests there

were 1.3 million disabling injuries caused by hospital medical treatment. Moreover, this includes

173,000 adverse event deaths, including 90,000 deaths due to medical negligence. Both of these

would rank among the leading causes of death in the US: adverse event deaths would be the third

and medical negligence deaths the sixth leading cause of death.6

4 An injury rate of 4 percent may sound small. However, if people had, on a day-to-day basis, a 4 percent chance of

injury, they would be expected to face serious disabling harm every 25 days.

6 Here I update calculations done by the Institute of Medicine originally using 1997 data (Kohn et al 2000:31). Data on hospital

admissions and leading causes of death comes from the National Center for Health Statistics

(http://www.cdc.gov/nchs/fastats/Default.htm). The calculation proceeds as follows: 3.7% of admissions result in an adverse

event, and 13.6% of adverse events result in death. Brennan et al (1991:371-72) report that 51.3% of adverse event deaths were

due to medical negligence (higher than the overall rate of negligence). The Institute of Medicine (Kohn et al 2000:31) calculates

the number of adverse event deaths due to medical error (rather than negligence), using the estimate that 58% of all adverse

events involved error. However, the rate of error among adverse event deaths is not reported in the Harvard study; negligence

(and presumably error as well) is much more frequent in cases of severe injury, so the estimate by the Institute of Health is

probably too low. Using their formula generates an estimate of 101,000 deaths due to medical error in 2004.

11

Given the risks involved, one might expect patients to be highly concerned about the

quality of the medical care they are receiving. However, a number of factors make medical

quality difficult for patients to observe. The classical theory of Arrow (1963) and Parsons (1951)

emphasizes the complexity of medical care and the uncertainty of treatment outcomes as core

reasons why patients have limited ability to monitor their treatment. While this is, no doubt, a

fundamental grounding, we also emphasize the culture of medical practice. Physicians manage

patients‟ expectations through displays of confidence, alternating usage of pessimism and

optimism, and the non-disclosure of mistakes.

The complexity of medical care is an obvious barrier to patient oversight. Much of

medicine happens “behind the scenes,” where patients are either not awake or not directly

involved. For example, when doctors judge the surgical skill of other doctors, they do not rely on

bedside manner, friendliness, or technical chatter. Among doctors, signs of surgical skill are

things like being able to tie a “one-handed surgeon‟s knot” (which indicates “good hands”)

(Cassell 1991:11-12). “Judgment” is something that surgeons also emphasize as the mark of a

great surgeon. This can really only be observed in the course of surgery – seeing how a doctor

manages unexpected problems, challenges, and crises in the operating room. As one surgeon

remarked, “everything boils down to the time of surgery – what you did during surgery. The rest

is window dressing” (quoted in Cassell 1991:29). Indeed, a doctor‟s willingness to have a

prolonged discussion with a patient may be a negative signal – a sign that the physician has a low

opportunity cost of time (ie, not many surgeries to perform). In any event, when surgeons try to

gauge the ability of other surgeons, they do not use „conversation with patients‟ as a test of

surgical skill. Patients, however, often have little more to go on.

12

This combines with a culture of medical practice that trains doctors to project confidence

and certitude (Fox 1957; Light 1979). In medical training, young surgeons are taught to be

decisive, to show confidence in their decisions; they are criticized for showing too much

uncertainty and are taught that “doubt makes them less effective with patients” (Cassell

1991:57). Surgeons also believe that instilling confidence and optimism in patients aids in their

recovery. The typical post-operative patient feels extraordinarily bad; confidence in their surgeon

can have a strong placebo effect. Patients who believe in their surgeons (and thus in the

beneficial effects of their surgery) have less anxiety and remain in better spirits, manage their

post-operative pain better, and move more quickly into physical therapy and rehabilitation. For

these reasons, surgeons often feel that projecting confidence is part of their role as a good

doctor. As one surgeon said, “you‟ve got to make patients feel that you think their operation isn‟t

all that serious, that you know you can handle the operation, and that they‟re going to get better

fast” (quoted in Cassell 1991:23).

Physicians also tend to invoke a combination of optimism and pessimism to motivate

treatment and even to influence the course of recovery (Christakis 1999). In diagnosis,

physicians are pessimistic by nature: “when in doubt, suspect illness.” This pessimism enforces a

strict vigilance that ensures no symptoms are ignored or treated too lightly – that doctors always

err on the side of caution. However, in prognosis – the forecasting of patient outcomes –

physicians embrace an optimistic bias. This ensures that doctors do not give up on a patient “too

soon”. Physicians, in short, are prone to see illness everywhere, but to retain hope of recovery no

matter how bleak the situation.

The optimistic belief in recovery is in part because doctors feel that their predictions can

“take on a life of their own” (Christakis 1999: quoted, p156). To tell a patient that she is dying,

13

or that she will never walk again, or that she will not make a complete recovery is to undermine

the patient‟s determination to survive and improve. Patients may become depressed, stop eating,

disengage from treatment, and begin wasting. In severely ill patients, doctors see prognosis as

like “handling a bomb”: what they tell patients can shape the future and alter the course and

timeline of disease. Displays of optimism are often at their highest when the medical case is most

uncertain. Optimism is an appeal to patient will-power, and doctors draw on this resource most

when there is a very wide range of possible medical outcomes and when they are unsure

“whether their technical or scientific know-how is adequate to the challenge” (Christakis 1999:

quoted, p166). In these cases, doctors implore optimism in their patients, trying to bring patient

will-power onto the side of an uncertain medical course.

These factors make doctors very difficult to read. It means that physicians are not simply

providing objective information to the “patient-as-decision-maker”. Rather, they enrich and color

the information they present with the goal of guiding treatment choices, and maximizing credible

hope and determination in patients.

Moreover, when mistakes are made during the course of medical treatment, physicians

are very reluctant to discuss these errors with patients (Mazor et al 2004; Studdert et al 2007;

Wu, et al 1991). While doctors express support for the principle of full disclosure, it seems in

practice only about 25 percent of the time do physicians discuss their substantive mistakes with

patients (Mazor et al 2004). And when they do, such disclosures are carefully worded

descriptions of unexpected challenges that avoid attributions of responsibility (ibid). When

mistakes happen, doctors typically take cover behind the complexity and uncertainty of

medicine. This is sharply different from the tone of internal physician review. “Morbidity and

Mortality Conferences” at academic hospitals provide weekly, closed-door review of mistakes,

14

mishaps, injuries and deaths. Cases are vigorously dissected, responsibility attributed, and

solution (the correct course of treatment) is explicitly laid out. The philosophy of internal review

is that every bad outcome should have been avoided by doctors taking a better course of action

(Gawande 2002). This contrast between vigorous internal review and non-disclosure to patients

vividly captures the issue of asymmetric information. When medicine goes wrong, doctors

themselves carefully analyze what errors were made and by whom; patients, however, are told

only that treatment did not go as well as hoped.

In summary, doctors are taught to instill confidence in their patients in order to minimize

anxiety, sustain hope, and facilitate belief in recovery. When treatment goes awry, doctors

emphasize the difficulties of their ailment and the uncertainties of medical outcomes. Faced with

these consistent messages, it is difficult for patients to sense when they are receiving less-than-

ideal clinical care.

Existing Evidence on Patients’ Ability to Detect Quality

In this section we review existing research on patient knowledge about the quality of

medical care. One empirical strategy focuses on malpractice lawsuits. The threat of malpractice

litigation encourages medical providers to offer appropriate, diligent care. However, such

litigation typically depends on the patient‟s ability to identify an episode of dubious or negligent

treatment. The Harvard Medical Practice Study sheds some key light on this issue. Most lawsuits

did not involve an injury caused by medical treatment. These were medical cases that had a poor

outcome, but one caused by the underlying disease or injury rather than caused by the medical

treatment. Moreover, among the several hundred cases that the Harvard Study flagged as

negligent care, only a small handful (1.5 percent) became lawsuits (Localio et al 1991). A similar

15

study of Colorado and Utah likewise found that negligent adverse events led to litigation in only

2.5% of cases (Studdert et al 2000). This is characteristic of an information problem: negligent

medical care faces almost the same rate of malpractice lawsuits as non-negligent care. Negligent

care does experience a higher rate of litigation (1.5%) than non-negligent adverse events (0.8%)

(Localio et al 1991:248). However, the difference is minimal and cannot serve as a major

deterrent for a rational physician who is contemplating the costs and benefits of providing

sloppy, inattentive, or negligent care. For example, if negligent care was litigated 50 percent of

the time, or even 20 percent of the time, this would be a powerful force ensuring adherence to the

proper standard of care. At the current rates, litigation is a background annoyance for doctors

almost independent of the quality of the care they provide.

There are, no doubt, a number of factors that deter malpractice litigation. Patients, or their

surviving relatives, are averse to legal conflict and want to get on with their lives. Lawyers

themselves, working on contingency fee basis, effectively ration the number of lawsuits and only

accept the most promising legal cases. Yet, with such a low rate of litigation even in the case of

negligence, it seems clear that most patients simply do not know when they have received

negligent care. Studdert et al (2007) estimate that full disclosure of errors to patients would lead

to a substantial increase in malpractice litigation.

Another approach to analyzing asymmetric information in medicine focuses on the

relationship between clinical quality and patient satisfaction. The existing studies show

somewhat mixed results. A Canadian study noted that medical quality was highest at emergency

rooms, lower in walk-in clinics, and lowest at family practices. However, patient satisfaction

rankings were completely reversed: top rated were family practices, followed by clinics, and

lowest at emergency rooms (Hutchison et al 2003). In short, patients were inclined to

16

recommend, and return to, lower-quality medical providers. Lee et al (2008) followed nearly

2,000 heart attack patients. Satisfaction was high overall (92 percent satisfied), regardless of

technical quality of treatment. Moreover, satisfaction with care did not predict long-term survival

or the probability of recurrent heart attack. A small study of elderly patients (Chang et al 2006)

found that technical care had no association with patient satisfaction ratings. However, quality of

interpersonal interaction with medical staff had a very strong correlation with patients‟ overall

assessment of the quality of their care (Chang et al 2006). All of these studies suggest that

patients are not well aware of the tradeoff between technical medical quality and hospitality.

On the other hand, several studies report contrary findings. In a study of hospitals in

Cleveland, hospitals with lower mortality rates had higher levels of patient satisfaction (Jaipaul

and Rosenthal 2003). In another study, Jha et al (2008) found that “care was consistently better

in the hospitals that received high [patient] ratings” and concluded that “there is no need for

tradeoffs between” technical quality and patient satisfaction (p. 1930). In pediatric care, Schempf

et al (2007) found that parent dissatisfaction was a reliable marker of an inappropriate course of

preventative child medicine. Finally, Rodriguez et al (2007) studied factors that lead HIV

patients to change doctors. Patients were more likely to switch away from doctors that tested

poorly in antiretroviral knowledge. As the authors concluded, “these findings challenge the

notion… that patients are unable to assess the technical quality of care they receive” (Rodriguez

et al 2007: 194). In short, the existing studies offer mixed evidence on the degree of information

problems in medicine. Moreover, existing studies do not directly compare the effects of

hospitality and clinical quality on patient satisfaction7, which is central to our focus on the partial

information problem.

7 The exception is Chang et al (2006), who study 230 elderly patients.

17

This study examines patient inferences about the quality of hospital care. We draw on a

large sample of US hospitals, combining data on technical medical quality, patient survival rates,

hospitality aspects of care, and patient satisfaction. We test how various medical and non-

medical aspects of hospital treatment influence patient satisfaction and willingness to

recommend the hospital to others. Is patient satisfaction higher in hospitals with top notch

medical quality?

Our hypotheses are as follows:

1) patients have little ability to infer the technical quality of their medical care;

2) patients have vivid knowledge of the “room and board” aspect of their care; and,

3) overall evaluations of hospital care are driven by what patients can observe, so that

temporary hospitality trumps clinical quality in patients‟ own evaluation of their hospital.

This paper also tests whether the competitive context of hospital care matters: do more

competitive hospital markets facilitate information flow and better patient inferences about

quality? This is what one would expect if markets are able to solve their own information

problems, in which higher-quality producers find effective ways to signal their superior quality

to consumers (Spence 1974). For example, in more competitive environments, the best hospitals

may spend more effort advertising their technical quality ratings.

Data Set

Our data combines hospital-level information on patient satisfaction, technical medical

quality, patient mortality rates, and hospitality aspects of care. Some 2,684 hospitals (about 55

percent of all acute care hospitals in the US) are included, for the period 2006-07. Hospitals

voluntarily participate in the data collection program. Thus, it is a large sample of hospitals, but

not a random sample, and may not be representative of all hospitals in the US. Missing data on

18

different variables brings the final sample down to 2,160 (80 percent of the original sample).

These data were obtained from the Hospital Compare Database, provided by the Centers for

Medicaid and Medicare Services. Descriptive statistics for the full data set are provided in Table

1.

[Table 1: Descriptive Statistics]

Patient satisfaction. The outcome variable in this study is patient satisfaction. Data are

from the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS)

survey. Patients are asked whether they would recommend their hospital to friends and family,

and to give an overall rating of their hospital. This provides two complimentary measures of

patients‟ own assessments of the hospital care they experienced. The data are aggregated at the

hospital level, reported as the percentage of patients at each hospital giving a “high” rating (9 or

10 out of 10), moderate rating (7-8 out of 10), or low rating (0 to 6). It is clear from Table 1 that

patients are quite favorable to their hospitals; the modal response is a 9 or 10 out of 10, with 62

percent giving this high rating. Only 11 percent of patients seem clearly dissatisfied. Likewise,

67 percent say they would “definitely” recommend their hospital, while only 7 percent say they

would definitely NOT recommend.

We use this data to estimate patients‟ ability to detect the underlying quality of hospital

clinical care. Patients should be more satisfied with their hospital, not only when hospitality

measures rank high, but also when technical quality is higher, and when hospital survival rates

are higher.

Technical Medical Quality. Our data on medical quality are based on adherence to the

standards of care for heart attack, heart failure, pneumonia, and general surgical practice.

Measures were selected by the National Quality Forum, an independent advisory board made up

19

of doctors, nurses, hospital administrators, and other stakeholders. The 24 measures of technical

medical quality were selected for their relevance to health outcomes, reliable measurability, and

need for national improvement in medical practice. The basic need for measurability, in

particular, means that the items cannot cover all aspects of technical medical quality. For

example, the measures do not indicate how well medical teams respond when surgery goes

wrong, when some unexpected rupture suddenly turns a routine operation into a crisis. The data

do not indicate how hospital staff made a heroic diagnosis from a deeply puzzling case, or caught

the subtle symptoms of a dangerous infectious disease before it was “too late”. The data do,

however, provide important indicators of the hospital medical environment – how swiftly and

reliably they act to treat acute illness and uphold patient safety.

For heart attack care, the measures record whether and how quickly patients are given

medication to dissolve blood clots or reduce blood pressure. If coronary surgery is needed, is it

performed within two hours of admission? These measures focus on whether and how quickly

action is taken to open blocked coronary arteries.8 For pneumonia, the measures focus on the

timeliness of treating with antibiotics, whether blood tests were taken prior to administering

antibiotics, whether the patient‟s blood oxygen level was evaluated, and whether the most

appropriate antibiotic was selected. Measures of surgical care focus on the prevention of

infection, and the appropriate use and selection of preventative antibiotics. For heart failure,

measures include whether a test was given for how well the heart is pumping blood (e.g.,

electrocardiogram, chest x-ray) and whether proper medication was given in the case of heart

dysfunction.

8 Note that some patients are contraindicated for standard treatments. When patients are not given treatments for which they are contraindicated, this is appropriately recorded in the dataset.

20

The full list of quality measures appears in Appendix I. In our data set, hospitals provide

appropriate care 86 percent of the time. In 14 percent of hospitals, the appropriate standard of

care was not applied. Many of these quality measures are being incorporated into new operating

room checklists in an effort to ensure that the fundamentals are done correctly every time,

without error (Gawande 2009).

Patient survival. Mortality / patient survival data looks at whether acute care Medicare

patients died within 30 days of their hospital admission. The measure includes patients initially

admitted for heart attack and heart failure. The mortality rates are also “severity-adjusted” to

control for how sick patients were at their time of admission. For example, patients with more

severe symptoms, a history of heart disease, who are older and arrive in the hospital with co-

morbidities such as diabetes, malnutrition, or liver disease, are more likely to die regardless of

the quality of medical care. Severity-adjusted mortality is generated from a hierarchical

regression model that controls for these individual risk characteristics. Essentially, this compares

a hospital‟s death rate conditional on its mix of patients, compared with the expected death rate

of the “average hospital” (given the same mix of patients). Such a procedure identifies hospital-

specific quality – whether the hospital has a better or worse death rate given its mix of patients.9

Medicare, however, does not release the actual mortality rates, but rather publishes the

results of the significance test: is a given hospital‟s mortality rate significantly higher,

significantly lower, or statistically equivalent to the national average? In short, Medicare

publishes the “stars and signs” from the regression output but does not report the actual estimates

themselves. Moreover, there are evidently large standard errors around the estimates, as very

few hospitals achieve a significantly better or worse death rate than average. Because of the

9 For more information, see Medicare‟s information for professionals website: http://www.hospitalcompare.hhs.gov/staticpages/for-

professionals/ooc/calculation-of-30-day-risk.aspx

21

small number of hospitals with significant differences, we use this as supplemental data to

compare the very “best” and “worst” hospitals in our data set.

Hospitality. The “room and board” aspects of hospital care are measured from a battery

of items in the HCAPS patient survey. Did the nurses treat patients with courtesy and respect,

listen carefully to them, and explain things in ways patients can understand? Did patients feel

their pain was well controlled, and did staff do “everything they could” to help with pain

management? When patients pressed the call button, did help arrive quickly (“as soon as you

wanted it”)? Did hospital staff tell patients what their medication was for, and did they explain

possible side effects of medication? Were patients told “what to expect during their recovery at

home”? Were the rooms kept clean? Were the rooms quiet at night?

Like the mortality data, patient reports of room and board quality are adjusted for the

patient mix in each hospital. For example, older patients, those in poor health, and those with less

education tend to be more critical of the hospitality aspects of care. This may be because they are

more intrinsically dissatisfied, or because such patients actually tend to receive less respectful

and considerate care. Either way, the average effects of differences in patient mix between

hospitals are stripped out using regression adjustment. Hence, the differences show whether a

nationally representative patient is reporting better room and board care in a given hospital.

Market Competition. To obtain data on the competitive environment facing hospitals, we

merged in data from the Health Care Cost and Utilization Project, using the latest (2003) hospital

market structure files. Competition, therefore, is measured with a three year lag relative to other

hospital characteristics. We expect that the degree of local market competition changes slowly

over time, so that the 2003 data still provides a relatively good measure of competition in 2006-

07. Inevitably, however, the lag will introduce noise that will bias the regression coefficients

22

towards zero, making this a conservative test of the effect of competition. Moreover, matching

hospitals across data sets proved difficult, had to be done by hand, and only 445 hospitals could

be matched with competition data. Thus, testing the effect of competition can only be done using

a sub-set of the data. One upside is that we know which hospitals have missing data on

competition, and can reweight this smaller data set so that it matches the broader characteristics

of our full sample. An additional test of representativeness is whether the main results can be

replicated on the sub-sample.

Market competition is a classic case of model uncertainty: there are 18 different ways of

estimating the degree of local market competition (the presence of other hospitals that may draw

away patients). This includes different ways of defining a local market (political boundaries,

fixed radius, variable radius, and patient flow) and different measures of the intensity of

competition within the local area (number of hospitals, Herfindahl index) (Wong, Zhan, and

Mutter, 2005). The Herfindahl index looks at the market share of each hospital in a market, with

a metric ranging from zero (perfect competition / many small competitors) to one (complete

monopoly of one firm). The other prominent approach assumes that the specific market shares

are less important than simply the number of hospitals competing in a market. Each of these two

approaches comes with nine different ways of defining a local market. Rather than trying to

select one or two preferred measures, we use all measures, testing them one at a time, and

considering the weight of the evidence.

23

Methods

We estimate the relationship between the quality of medical care and patients‟

satisfaction with and willingness to recommend their hospital. With two outcome variables, we

have two equations:

Satisfactioni = + Qualityi + Z + X

+ υi (1)

Recommendationi = + Qualityi + Z + X

+ i (2)

In each model the subscript i denotes the hospital. Qualityi represents the technical quality of

medical care. Z is a vector of variables capturing the “room and board” quality of hospitals. X is

a vector of hospital- and state-level control variables. The terms υi and i are random

disturbances associated with the respective outcome variables.

Given that both equations include the same set of right-hand side variables, they can be

pooled and estimated jointly within one model.10

Pooled regression is similar to a panel study in

which hospitals are observed at two different time periods; in this case, however, hospitals are

observed on two similar outcomes at one time. Technically, this creates a hierarchical data

structure in which observations are nested within hospitals. As the error terms (υi and i) are

likely to be correlated within hospitals (which generates misleading standard errors), we use the

random effects model – generalized least squares – rather than pooled OLS. Writing the above

two equations as one jointly-estimated model,

Yi = +αQualityi + Z + X + i (3)

10

Another approach could be to simply average the two outcome variables, although this has the effect of reducing

the amount of analyzable information. The pooled analysis preserves the full information, allowing analysis of

differences between outcome variables (ie, differences across questions).

24

where the subscript o denotes the specific outcome; when o=1, the outcome is patient

satisfaction, and when o=2 the outcome is patient recommendation.

An additional feature of our data is that the outcome variables are not fully continuous.

Rather, they are percentages or fractions bounded between zero and one: such as the percent of

patients reporting “high” satisfaction, or the percent of patients “definitely” recommending their

hospital. This is similar in principle to a problem of binary outcome variables. As Papke and

Wooldridge (1996:620) note, “the drawbacks of linear models for fractional data are analogous

to the drawbacks of the linear probability model for binary data”. OLS / GLS are still consistent

estimators (ie, unbiased in large samples), although as Kieschnick and McCullough (2003) and

Paolino (2001) emphasize, Beta Maximum Likelihood Estimation is likely to be a more efficient

estimator. The Beta MLE model is written as,

)1

ln(

δ + Qualityi + Z + X (E(y|x) = ) (4)

where 0<y<1, 0<μ<1, and μ is the conditional mean value of y.11

In a sense, this is a logistic

regression where the dependent variable is a proportion. This is our preferred model, although in

11

Our model here assumes dependent variable Yi has a beta distribution and its standard density function is

11 )1()()(

)(),|(

yyyf , where 0<y<1, α,β>0, Г() denotes the gamma function. About the beta

regression model in details and literature, see Kieschnick and McCullogh (2003), Paolino (2001), and Smithson and

Verkuilen (2006). We use alternative parameterization expressed as: 11 )1()()(

)(),|(

yyyf . E(y)=

μ,

1

1)1()(yVar , where 0<y<1, µ,ϕ>0, Г() denotes the gamma function. In this article, we restrict

attention to the mean model (μ) or the location submodel (term used by Smithson and Verkuilen (2006)). We use

Stata command betafit in getting our estimations of equation 4. Note that to meet the condition that 0<y<1, thirty

observations of 0 were recoded as 0.01.

25

substantive terms, the results from the linear specification (equation 3 above) are little different,

and primarily have larger standard errors reflecting the lower efficiency of the linear model.

Results from the linear models are available from the authors.

Finally, because of the unique structure of the data, we must separately analyze high and

low levels of patient satisfaction and willingness to recommend. Patient satisfaction is

represented by three variables: the percent with “high” satisfaction” (9-10 out of 10), the percent

with “medium” satisfaction (7 – 8 out of 10), and the percent with “low” satisfaction (0 to 6 out

of 10). These variables sum to one for each hospital, and we use high and low satisfaction to

analyze all the informative variation.12

This is a byproduct of hospital-level, rather than

individual-level, measurement. This does, however, give a natural way of testing whether

positive evaluations are generated by the same basic process as negative evaluations. We report

these as positive response models and negative response models. If the same processes that

produce positive responses are also generating negative responses, the negative response models

should yield approximately the same results as the positive response models, but with opposite-

signed coefficients. Negative responses may have special significance, as strong negative

evaluations of a hospital may indicate willingness or desire to litigate.

Findings

Main Results: Technical Medical Care and Hospitality

Table 2 shows the positive response models, which estimate the determinants of a

hospital receiving positive evaluations from a larger proportion of patients. Model 1 shows the

bivariate relationship between patient satisfaction and the technical quality of medical care. This

12

Studying variation in moderate satisfaction is redundant, as the values for moderate satisfaction are fully

determined by the values of the other two variables.

26

shows that patient satisfaction and medical quality move together in the same direction. Model 2

adds in hospitality variables as well as hospital-level and state-level controls. The effect of

technical quality is reduced but remains positive and significant. The elasticity (0.09) indicates

that a 10 percent increase in medical quality leads to a 0.9 percent increase in patient

satisfaction.13

Model 3 uses patient recommendation (the percentage definitely recommending

their hospital) rather than patient satisfaction, and the results are quite similar. Model 4 pools

these two outcome variables, giving an overall elasticity of “satisfaction” with respect to medical

quality of 0.11: a 10 percent increase in medical quality leads to a 1.1 percent increase in positive

evaluations. Almost all of the hospitality variables have estimated effects larger in magnitude

than medical quality. Cleaner and quieter rooms each lead to greater increments in positive

responses (elasticities of 0.14 and 0.18 respectively) than does medical quality. Better pain

management and more effectively communicating information about recovery at home have

even larger effect sizes. The largest factor, however, is the quality of nurse communication.

Improving nurse communication by 10 percent leads to a 6.6 percent increase in positive

evaluations – an effect size some six times the magnitude of medical quality.

[Table 2: Fractional Regression: Positive Response Models]

[Figure 1: Predicted Effects of Medical Quality and Nurse Communication on

Overall Satisfaction]

Table 3 reports on the negative response models, the likelihood of hospitals receiving

negative evaluations from a larger proportion of patients. Recall that if the same causal processes

13

Additional results (not reported) show that patients are most sensitive to quality in heart attack treatment, and

show less responsiveness to the quality of care for heart failure, pneumonia, or general surgery.

27

generate both positive and negative patient responses, then coefficients in the negative response

models should have the opposite signs as in the positive response models.

The technical quality of medical care never achieves statistical significance as a

determinant of negative patient responses (models 5 through 8). In the bivariate relationship with

negative satisfaction, medical care has a relatively large elasticity (0.22), but this falls to zero

when controls are added (model 6). For negative recommendations (model 7), the effect of

medical care is small and non-significant. In the pooled model, increasing the quality of medical

care by 10 percent leads to a 0.6 percent lower risk of negative patient evaluations. In short,

when patients complain about their hospitals, the technical quality of medical care has little to do

with their concerns.

The hospitality aspects of care have even greater importance when it comes to negative

patient responses. Improvements in pain management, information about home-based recovery,

and room cleanliness all have much larger elasticities than they do in the positive response

models. And the quality of nurse communication is the over-riding factor in generating or

discouraging negative hospital ratings. A 10 percent increase in nurse communication leads to a

striking 23.1 percent reduction in negative patient responses. When patients complain about their

hospitals, it is due in large part to how the nurses treated them.

[Table 3: Fractional Regression: Negative Response Models]

It should be noted that two aspects of hospitality have small and “wrong-signed”

coefficients in both the positive and negative response models. The general responsiveness of

hospital staff unexpectedly turns out with a negative coefficient in the positive response models

(and a positive coefficient in the negative response models), albeit small and only modestly

significant. Staff communication about medicines is also wrong-signed but small and non-

28

significant in almost all models. Our interpretation is that these factors, after conditioning on the

other aspects of hospitality, are not very important to patients.

Overall, the main conclusion is that per-unit increases in the hospitality aspect of care

have a much greater effect on patient satisfaction than per-unit increases in technical medical

quality (see figure 1, which graphs the effects on positive responses). Auxiliary regressions on

medical quality itself (not reported) reinforce this picture. For example, hospitals with 10 percent

higher quality of nurse communication have 1.5 percent higher medical quality, but 6.6 percent

higher patient satisfaction. Hospitality is not a negative signal for medical quality (bad hospitals

compensate by investing in hospitality), but it is a distracting signal for patients that leads to

exaggerated swings in evaluations of care. A profit maximizing hospital aiming to increase

patient satisfaction would likely look first to relatively superficial hospitality aspects of care.

This becomes an exercise in impression management –emphasis on cultivating an “emotional

experience” for patients.

Hospitals are rewarded by patients primarily for providing hospitality, and only

secondarily for providing excellent medical care. This does not necessarily reflect patient

priorities, but rather simply what patients can observe about their hospital stay.

Other hospital characteristics.

The above estimates include regression adjustment for differences in hospital ownership,

Medicare pricing, presence of emergency service, and survey response rate, as well as state-level

differences in education, per capita GDP, and population density. For purposes of space,

estimates for state-level controls are not reported. The results for hospital ownership indicate that

private hospitals have higher levels of patient satisfaction than government-owned hospitals (the

29

reference category), regardless of whether they are non-profit or for-profit. However, for-profit

hospitals also have significantly higher levels of negative evaluations (9.6 percent more patients

give negative ratings). Non-profit hospitals, in contrast, do not have significantly higher rates of

negative ratings. For-profit hospitals generate more extreme responses than government-owned

operations – both more “very positive” and more “very negative”. Non-profits seem to be the

Pareto-improving option, having higher positive scores but no greater negative scores as

government-owned hospitals. The presence of an emergency service at a hospital has no

significant effect on satisfaction.

The Medicare price paid to a hospital is positive and strongly significant, though the

interpretation of this variable is somewhat unclear. Medicare aims for horizontal equity in its

reimbursement rates – meaning that hospitals are paid the same prices for the same procedures.

There are, however, regional cost adjustments, which presumably accounts for much the

variation in Medicare prices (Reinhart 2006). This suggests that hospitals in areas with higher

cost of living enjoy higher patient satisfaction scores.

The coefficient for survey response rate is positive and significant, meaning that when

hospitals have higher response rates they tend to have higher patient satisfaction scores.

Willingness to respond to the patient survey seems to be based on how satisfied patients are with

their hospital.

Patient Survival Rates

As an additional test, Table 4 shows hospitals with the best and worst 30-day survival

rates for two types of treatment: heart attack care, and heart failure care.14

There are only a small

number of hospitals in the data set with survival rates that are far enough from the average for

14

Survival rates are adjusted for severity of illness and other risk attributes of patients.

30

Medicare to flag them as generating a remarkable number of deaths (40 are flagged as the best,

29 flagged as the worst). However, this allows a test of whether patients can tell the difference

between the very best and the very worst hospitals in the country, in terms of death rates for

heart failure and heart attack care. In other words, this is a “most-likely” empirical test design,

comparing hospitals with the most extreme differences in the most critical of medical outcomes.

Looking at positive responses, hospitals with the best survival rates have an average

patient satisfaction score of 65.2, compared to 64.0 at hospitals with the worst survival rates. The

difference, 1.2 percentage points, is in the expected direction but remarkably small and far from

statistically significant (even with relatively small standard errors). For negative responses, the

gap is even smaller: at hospitals with the best survival, 9.1 percent are explicitly dissatisfied with

their treatment, compared to 8.9 percent at hospitals with the worst survival rates (a difference of

0.2 percentage points). Even the most extreme differences in hospital-survival outcomes generate

negligible differences in patient satisfaction.

[Table 4: Hospital Survival Rates and Patient Satisfaction]

Market Competition

As discussed above, the measures of market competition were merged in from a different

data set, which yielded roughly 450 hospitals with matching data on competition, medical

quality, and patient satisfaction. There are 18 measures of competition, and three outcome

variables (positive response, negative response, and medical quality). Table 5 shows the key

coefficients of interest from 54 regression models.

For positive satisfaction responses, the signs on competition are positive in 17 out of 18

measures, and significantly so in 13. The weight of the evidence strongly supports that

31

competition increases patient satisfaction. Looking at negative responses, when people are

explicitly unhappy with their hospital, the signs indicate that competition reduces patient

discontent for 16 measures, though the coefficients are small and only significant for three

measures. This suggests that competition may reduce patient dissatisfaction, but the effect is

probably too small to matter.

For medical quality, the results are somewhat more contingent on the specific measure.

The signs on competition are negative in 14 out of 18 coefficients, and significantly so for 10 of

those. There are two measures that show statistically significant opposite signs. There is less

certainty about this result than for the effect of competition on positive responses. However, the

weight of the evidence indicates that medical quality is lower when there is greater hospital

competition.

[Table 5: Effect of Competition on Patient Satisfaction and Medical Technical

Quality]

Additional results (not reported) show that in the subset of data for which competition data are

available, the baseline results – the effects of medical quality and hospitality – are very similar to

those reported above in Tables 2 and 3. This suggests that the smaller sample is representative of

the full data set. Moreover, interaction effect of competition and medical quality has a clear zero

coefficient, indicating that patients‟ ability to identify the quality of their hospital does not

depend on the level of market competition. In other words, more intense market competition

does not improve the flow of information in hospital markets, nor lead the best hospitals to more

effectively signal their quality to patients.

32

Discussion

Hospitals balance two aspects of patient care. First is technical medical quality, which

represents the reason why patients are under their care. The second is hospitality treatment –

maintaining patient comfort during their stay. This is tangential to patients‟ long-term well-

being, but is a visible and memorable aspect of their hospital experience.

Drawing on a sample of over 2,000 American hospitals, this research finds that patients

have very limited ability to observe the technical quality of their medical care, but are much

more sensitive to the quality of room and board care. Raising medical quality by 10 percent leads

to only a 1.1 percent increase in positive reports of satisfaction or in willingness to recommend

the hospital. When patients are explicitly unhappy with their hospital, the quality of medical care

is completely unrelated. In contrast, the hospitality/„room and board‟ care is a key driver of

patient satisfaction. The quality of interaction with nurses has an effect size (elasticity) some six

times larger than medical quality. Even relatively minor customer service aspects, such as the

quietness of rooms, have more impact on patient satisfaction than does medical quality.

Moreover, hospitals with the very highest death rates in the country (for of heart attack

and heart failure care) receive the same patient satisfaction scores as hospitals with the very

lowest death rates. Patients do not distinguish between the very best and very worst hospitals on

these dimensions.

33

Crowding Out Medical Care

Patients face an acute problem of partial information in the market for hospital care,

where the most important aspect of quality is hardest for patients to observe. This carries great

potential to distract both patients and hospitals from the core mission of medical excellence. In a

medical market with more high-charged incentives, competition for patients may lead hospitals

to focus on what their consumers can observe, and skimp on what they cannot. In a truly market

driven (or “consumer driven”) health care market, one might expect to see developments such as

24-hour room service, restaurant-quality meals, HBO channels, non-medical staff to tend to

patient comfort, and hospital executives recruited from the service industry. These are not bad in

their own right, and certainly there are hospitals in desperate need of better „customer service‟

(see Cleary 2003). No doubt, patients suffering through the side-effects of medical intervention

will greatly appreciate a higher standard of hospitality. However, this same movement may lead

to cutbacks in what their customers cannot readily observe: the provision of excellent medical

treatment. Over time, hospitals may become increasingly comfortable places to stay, but less-

than-ideal places to undergo medical treatment. It is a market driven health care that turns

hospitals into hotels (Goldman and Romley 2008).

This is the theme of a recent award winning book, If Disney Ran Your Hospital: 9 ½

Things You Would Do Differently (Lee 2004). Hospitals, the author argues, must recognize that,

like Disney, they are providing an “emotional experience”. In this, perceptions are more

important than reality, and the perceived experience of the visit is more important than the

medical services provided. Drawing on the principles of a Disney production, Lee focuses on

how hospitals can cultivate a competitive advantage in hospitality.15

15

If Disney Ran Your Hospital won the 2005 best book award from the American College of Healthcare Executives,

and claims to have sold over 250,000 copies.

34

As a business strategy, investing in improvements in hospitality and amenities likely

offers a higher return than investments in medical quality. Indeed, these results indicate that if

hospitality and medical care had the same per-unit costs, hospitality investments would generate

between two to six times more patient loyalty than would better medical care.

Our limited competition data supports this shift in market focus. Hospitals facing greater

market competition have lower levels of medical quality, but higher levels of patient satisfaction.

Moreover, an interaction analysis finds no evidence at all that greater competition improves

patients‟ ability to identify the medical quality of their hospital. More intensive competition in

hospital markets seems to offer dubious benefits to patients.

One important caveat should be made. This is a study of acute care treatment; in areas

such as chronic disease management (diabetes, hemophilia, HIV, epilepsy) patients develop

substantial expertise in their illness and treatment. Britain, for example, has developed an “expert

patient program” to help patients with chronic disease to develop self-management skills, and to

help doctors appreciate patient expertise. Studies of chronic disease management would likely

show much greater ability of patients to detect the quality of medical care, as in the study of HIV

patients by Rodriguez et al (2007) (see also Cleary [2003] for an insightful case study of

hemophilia treatment). And competition for patients with chronic illness would be much more

likely to generate better, rather than worse, medical quality. Directly comparing a sample of

chronically-ill persons with non-chronic patients would be valuable and informative research.

As another avenue for future research, an excellent design would be to add a patient

satisfaction survey to a chart study of hospital medical errors and negligence (akin to the Harvard

Medical Practice Study). This would link the satisfaction of individual patients to the quality of

35

the specific treatments they received. Can patients tell when they received negligent medical care

(compared to when excellent treatment nonetheless had a bad outcome)?

Future research should also examine whether patients with limited insurance (the

“catastrophic coverage” advocated by the consumer-driven health care movement) are any better

at detecting medical quality, or whether they place a lower value on hospitality care. This would

help clarify the tradeoffs between a limited insurance, market-driven health policy agenda, and a

Medicare-for-all prescription.

More broadly, the problem of partial information has relevance to other markets of

sociological interest. In early childcare centers, parents have little ability to observe the quality of

care their children receive during the day, but can observe, and may be misled, by aspects such

as the cheerfulness and brightness of the center. This may encourage market competition based

on cheerful, bright, and clean centers rather than on quality programming that stimulates early

childhood cognitive and social development.

In higher education, potential students visiting campuses can observe the quality of

dorms and cafeterias (room and board hospitality), but have little ability to directly gauge the

quality of instruction. This may lead competing universities to shift investment away from

educational resources, and encourage students to select universities based on temporary quality

of life rather than on the long-term benefits of the education they will receive.

Exploring markets where problems of partial information are salient – where there is a

potential disconnect between superficial attributes and underlying quality – seems a promising

venture for economic sociologists.

36

Table 1. Descriptive Statistics

Variable

N Mean s.d. Minimum Maximum

Dependent Variables

Overall ratings (9 or 10, high %) 2517 63.23 9.84 12 94

Overall ratings (7 or 8, medium %) 2517 25.85 5.49 5 54

Overall ratings (6 or lower, low %) 2517 10.92 6.00 0 75

Recommendation (yes, definitely, %) 2517 67.48 10.49 23 100

Recommendation (yes, probably, %) 2517 26.45 7.95 0 77

Recommendation (No, not, %) 2517 6.07 3.91 0 26

Quality of Medical Care Variables

Overall quality of care 2322 86.41 6.78 44.58 99.81

Quality of heart attack care 2410 86.70 11.92 0 100

Quality of heart failure care 2483 84.65 12.45 0 100

Quality of pneumonia care 2493 88.44 7.78 0 100

Quality of surgical care 2380 82.51 11.21 0 100

Other Satisfaction Factor Variables

Nurse communication 2517 72.94 6.99 36 98

Communication about medicine 2517 57.69 7.50 29 100

Giving information to patient recovery at home 2517 79.17 5.86 54 99

Responsiveness of hospital staff 2517 60.28 9.26 18 91

Pain management 2517 67.31 6.36 41 97

Quiet room 2517 53.52 10.29 27 89

Clean room 2517 68.01 8.54 32 95

Hospital Characteristics Variables

Price($)/1000 2222 11.65 3.45 3.22 32.85

Ownership

Government 2517 0.16 0.37 0 1

Nonprofit 2517 0.68 0.47 0 1

Profit 2517 0.16 0.37 0 1

Emergency service (yes=1) 2517 0.95 0.211 0 1

Market competition 332 1.82 11.68 -17.32 37.31

Response rate (%) 2517 35.10 12.83 3 100

State Characteristics Variables

Education (% of population with college and

higher degree)

2517 51.16 5.50 35.80 63.70

GDP Per Capita (logged) 2517 10.54 0.15 10.17 11.83

Population density (logged) 2517 4.88 0.99 0.10 8.11

37

Table 2: Fractional Regression: Positive Response Models Independent

Variables

Satisfaction

(% giving high rating)

Recommendation Pooled

(% def. recommend)

Model 1 Model 2 Model 3 Model 4

Coef. Elasticity Coef. Elasticity Coef. Elasticity Coef. Elasticity

Technical Quality Quality of medical

care

0.72*** 0.24 0.29*** 0.09 0.43*** 0.12 0.36*** 0.11

(0.13) (0.07) (0.19) (0.07)

Hospitality Nurse communication 2.26*** 0.61 2.98*** 0.70 2.62*** 0.66

(0.18) (0.25) (0.23)

Information about

recovery at home 1.02*** 0.30 1.39*** 0.36 1.20*** 0.33

(0.11) (0.15) (0.14)

Pain management 0.70*** 0.17 1.11*** 0.24 0.90*** 0.21

(0.16) (0.21) (0.14) Quiet room 0.94*** 0.18 1.07*** 0.18 1.00*** 0.18

(0.06) (0.09) (0.06)

Clean room 0.80*** 0.20 0.43*** 0.09 0.61*** 0.14

(0.09) (0.12) (0.08)

Responsiveness of

hospital staff 0.02 0.00 -0.48*** -0.09 -.23* -0.05

(0.10) (0.14) (0.09) Staff communication

about medicines -0.12 -0.03 -0.34* -0.06 -0.23 -0.05

(0.13) (0.17) (0.12)

Other Hospital

Characteristics

Ownership

Government-Owned

(reference category) … …

Private, non-profit 3.91** N/A 4.82** N/A 4.37*** N/A

(1.33) (1.78) (1.22)

Private, for-profit 5.36*** N/A 6.17** N/A 5.75*** N/A

(1.66) (2.22) (1.52) Price ($) / 1000 2.16*** 0.09 3.47*** 0.13 2.81*** 0.12

(0.15) (0.20) (0.14)

Emergency service 1.04 N/A 2.19 N/A 1.66 N/A

(2.46) (3.30) (2.25) Survey response rate 0.70*** 0.09 0.95*** 0.11 0.82*** 0.10

(0.05) (0.07) (0.05) State-Level Controls

Included?

N Y Y Y

Constant -12.03 -277.56*** -361.15*** -318.67*** (11.17) (42.69) (56.72) (38.80)

Wald chi2 31.52 5,822.61 3,977.62 7,673.10

Log likelihood 2,069.57 3,541.96 3,055.05 6,115.18

Observations 2160 2160 2160 4,320

Notes: *p.05, **p.01, ***p.001 (two-tailed tests).

38

Table 3: Fractional Regression: Negative Response Models Independent

Variables

Negative

Satisfaction

Negative Pooled

Recommendation

Model 5 Model 6 Model 7 Model 8

Coef. Elasticity Coef. Elasticity Coef. Elasticity Coef. Elasticity

Technical Quality Quality of medical care -0.29 -0.22 -.002 -0.001 -0.11 -0.09 -0.07 -0.06

(0.17) (0.07) (0.12) (0.10)

Hospitality Nurse communication -3.32*** -2.14 -3.92*** -2.66 -3.50*** -2.31

(0.23) (0.29) (0.26) Information about

recovery at home -1.62*** -1.14 -1.64*** -1.22 -1.56*** -1.13

(0.14) (0.18) (0.16) Pain management -1.00*** -0.60 -0.88*** -0.55 -0.93*** -0.57

(0.14) (0.24) (0.22)

Quiet room -0.62*** -0.29 0.06 0.03 -0.25** -0.12

(0.08) (0.11) (0.10) Clean room -0.97*** -0.58 -0.94*** -0.59 -0.92*** -0.56

(0.08) (0.14) (0.13) Responsiveness of

hospital staff 0.58*** 0.31 0.16 0.09 0.34* 0.19

(0.13) (0.17) (0.15) Staff communication

about medicines -0.14* -0.07 0.03 0.02 -0.05 -0.03

(0.16) (0.21) (0.18)

Hospital Characteristics Ownership Government-Owned

(reference category) … … …

Private, non-profit -0.32 N/A 0.61 N/A 1.02 N/A

(1.75) (2.31) (2.00) Private, for-profit 3.59 N/A 16.07*** N/A 9.64*** N/A

(2.09) (2.70) (2.38) Price ($) / 1000 -1.55*** -0.16 -1.75*** -0.19 -1.60*** -0.17

(0.19) (0.25) (0.21) Emergency service -1.10 N/A 0.58 N/A -0.46 N/A

(3.16) (4.10) (3.60) Survey response rate -1.47*** -0.45 -1.66*** -0.54 -1.48*** -0.46

(0.07) (0.09) (0.08) State-Level Controls

Included?

N Y Y Y

Constant -12.03 331.58*** 352.67*** 329.32*** (11.17) (52.34) (66.26) (59.18)

Wald chi2 31.52 7,204.23 5,594.21 5,852.46

Log likelihood 2,069.57 4,954.82 5,549.53 8,859.18

Observations 2,160 2,160 2,160 4,320

Notes: *p.05, **p.01, ***p.001 (two-tailed tests).

39

Figure 1: Predicted Effects of Medical Quality and Nurse Communication on Overall

Satisfaction

55

60

65

70

75

Pre

dic

ted

Ove

rall S

atisfa

ctio

n (

%)

50 60 70 80 90 100 Score

Medical Quality Nurse Communication

40

Table 4. Hospital Survival Rates and Patient Satisfaction

Best

Survival

Rates

Worst

Survival

Rates Difference P-value

Number of Hospitals 40 29

Positive Responses

Patient Satisfaction 62.1 61.0 1.1 0.41

Std Error 1.5 1.8

Recommendation 68.3 67.0 1.3 0.82

Std Error 1.7 1.9

Overall Positive 65.2 64.0 1.2 0.66

Std Error 1.2 1.4

Negative Responses

Patient Satisfaction 11.8 11.7 0.1 0.56

Std Error 0.9 1.2

Recommendation 6.5 6.1 0.4 0.84

Std Error 0.7 0.9

Overall Negative 9.1 8.9 0.2 0.56

Std Error 0.6 0.8

41

Table 5: Effect of Competition on Patient Satisfaction and Medical Technical Quality

(Fractional Regression Coefficients)

Positive

response

Negative

response

Medical

quality

Herfindahl Index B

(SE)

B

(SE)

B

(SE)

Core-Based

Statistical Area

0.263*** -0.062 -0.150

(0.053) (0.086) (0.161)

County 0.248*** -0.047 -0.006

(0.050) (0.082) (0.130)

Health Service Area 0.298*** 0.018 -0.393^

(0.087) (0.145) (0.203)

Metropolitan

Statistical Area

0.125^ -0.017 -0.266

(0.075) (0.128) (0.220)

Fixed Radius 0.249*** -0.050 0.206

(0.047) (0.079) (0.127)

Variable Radius

75%

0.256*** -0.130* 0.304*

(041) (0.066) (0.121)

Variable Radius

90%

0.251*** -0.092 -0.086

(0.050) (0.081) (0.125)

Patient Flow 0.322** -0.007 -0.909**

(0.107) (0.173) (0.332)

Number of

Hospitals

Core-Based

Statistical Area

0.0003 -0.0006 -0.003**

(0.0004) (0.0006) (0.0009)

County 0.001^ -0.0008 -0.007***

(0.0007) (0.001) (0.002)

Health Service Area 0.001** -0.001 -0.004***

(0.0004) (0.0007) (0.001)

Metropolitan

Statistical Area

0.0003 -0.0005 -0.008***

(0.0006) (0.0009) (0.002)

Fixed Radius -0.0005 -0.0002 -0.005^

(0.0010) (0.0015) (0.003)

Variable Radius

75%

0.008*** -0.006* 0.0008

(0.002) (0.003) (0.0047)

Variable Radius

90%

0.0003 0.0001 0.001***

(0.0002) (0.0003) (0.0004)

Patient Flow 75% 0.015* -0.007 -0.068***

(0.007) (0.011) (0.020)

Patient Flow 90% 0.004 -0.005 -0.026***

(0.003) (0.004) (0.008)

Patient Flow 95% 0.004* -0.005^ -0.014**

(0.002) (0.003) (0.005)

Note: ^p.10, *p.05, **p.01, ***p.001 (two-tailed tests).

42

Notes to Table 5: The satisfaction models include all variables in main models (model 4 in Table

2 and model 8 in Table 3). The medical quality models include variables for all hospital

characteristics include teaching status, region, urban/rural, ownership, and bed size. To save

space, full results are not reported here, but available from the authors on request. The

Herfindahl Index is reverse coded (1 – index), so that larger values show greater (not lesser)

intensity of competition. This transformation only affects the signs of the coefficients.

43

References:

Akerlof, George. 1970. "The Market for 'Lemons': Quality Uncertainty and the Market

Mechanism". Quarterly Journal of Economics. Vol. 84(3): 488–500.

Arrow, Kenneth J. 1963. “Uncertainty and the Welfare Economics of Medical Care.” American

Economic Review. Vol. 53 (5): 941-973.

Berenson, Robert, and Christine Cassel. 2009. “Consumer-Driven Heath Care May Not Be What

Patients Need – Caveat Emptor.” Journal of the American Medical Association. Vol.

301(3): 321-3.

Brennan, Troyen, Lucian Leape, Nan Laird, et al. 1991. “Incidence of Adverse Events and

Negligence in Hospitalized Patients. Results from the Harvard Medical Practice Study I.”

New England Journal of Medicine. Vol. 324(6):370-76.

Brennan, Troyen, Colin Sox, and Helen Burstin. 1996. “Relation Between Negligent Adverse

Events and the Outcomes of Medical Malpractice Litigation.” New England Journal of

Medicine. Vol. 335(26): 1963-67.

Cassell, Joan. 1991. Expected Miracles: Surgeons at Work. Philadelphia: Temple University

Press.

Chang, John, Ron Hays, Paul Shekelle, Cathrine MacLean, David Solomon, David Reuben,

Carol Roth, Caren Kamberg, John Adams, Roy Young, and Neil Wenger. 2006.

“Patients‟ Global Ratings of Their Health Care Are Not Associated with the Technical

Quality of Their Care.” Annals of Internal Medicine. Vol. 144: 665-72.

Christakis, Nicholas. 1999. Death Foretold: Prophecy and Prognosis in Medical Care. Chicago:

University of Chicago Press.

44

Cleary, Paul. 2003. “A Hospitalization from Hell: A Patient‟s Perspective on Quality.” Annals of

Internal Medicine. Vol. 138(1):33-39.

Elwyn, Glyn, Stephen Buetow, Judith Hibbard, and Michel Wensing. 2007. “Respecting the

Subjective: Quality Measurement from the Patient‟s Perspective.” British Medical

Journal. Vol. 335: 1021-22.

Feldstein, Martin. 2005. “Rethinking Social Insurance.” American Economic Review. Vol. 95(1):

1–24.

Fox, Renee. 1957. “Training for Uncertainty.” Pp. 207-41 in Merton RK, Reader G, and Kendall

PL, (eds). The Student Physician. Cambridge, MA: Harvard University Press.

Gawande, Atul. 2002. Complications: A Surgeon’s Notes on an Imperfect Science. New York:

Picador.

Gawande, Atul. 2009b. The Checklist Manifesto: How to Get Things Right.

Goldin, Claudia, and Cecilia Rouse. 2000. “Orchestrating Impartiality: The Impact of "Blind"

Auditions on Female Musicians.” American Economic Review. Vol. 90(4):715–741.

Goldman, Dana, and John Romley. 2008. “Hospitals as Hotels: The Role of Patient Amenities in

Hospital Demand.” NBER Working Paper No. 14619.

Herzlinger, Regina. 1997. Market Driven Health Care: Who Wins, Who Loses In The

Transformation Of America's Largest Service Industry. Perseus Books.

Herzlinger, Regina (ed). 2004. Consumer-Driven Health Care: Implications for Providers,

Payers, and Policy Makers. San Francisco: John Wiley and Sons.

Herzlinger, Regina. 2007. Who Killed Health Care? America’s $2 Trillion Medical Problem --

and the Consumer-Driven Cure. New York: McGraw Hill.

45

Hutchison, Brian, Truls Ostbye, Jan Barnsley, Moira Stewart, Maria Mathews, Karen Campbell,

Eugene Vayda, Stewart Harris, Vicki Torrance-Rynard, Christine Tyrrell. 2003. “Patient

Satisfaction and Quality of Care in Walk-In Clinics, Family Practices, and Emergency

Departments: The Ontario Walk-In Clinic Study.” Canadian Medical Association

Journal. Vol. 168(8):977-83.

Jaipaul, Komal, and Gary Rosenthal. 2003. “Do Hospitals with Lower Mortality Have Higher

Patient Satisfaction? A Regional Analysis of Patients with Medical Diagnoses.”

American Journal of Medical Quality. Vol. 18(2):59-65.

Jha, Ashish, John Orav, Jie Zheng, and Arnold Epstein. 2008. “Patient Perceptions of Hospital

Care in the United States.” New England Journal of Medicine. Vol. 359(18):1921-31.

Kane, Robert, Matthew Maciejewski, and Michael Finch. 1997. “The Relationship of Patient

Satisfaction with Care and Clinical Outcomes.” Medical Care. Vol. 35(7):714-30.

Kieschnick, Robert and B. McCullough. 2003. “Regression Analysis of Variates Observed on

(0,1): Percentages, Proportions, and Fractions.” Statistical Modeling. Vol. 3:193–213.

Kohn, Linda T, Janet M. Corrigan, and Molla S. Donaldson (eds). 2000. To Err Is Human:

Building a Safer Health System. Washington, D.C.: Institute of Medicine, National

Academy Press.

Ladd, Helen. 2002. “School Vouchers: A Critical View.” Journal of Economic Perspectives.

Vol. 16(4):3-24.

Laine, Christine, Frank Davidoff, Charles Lewis, Eugene Nelson, Elizabeth Nelson, Ronald

Kessler, and Thomas Delbanco. 1996. “Important Elements of Outpatient Care: A

Comparison of Patients‟ and Physicians‟ Opinions.” Annals of Internal Medicine. Vol.

125(8):640-45.

46

Leape, Lucian, Troyen Brennan, Nan Laird, et al. 1991. “The Nature of Adverse Events in

Hospitalized Patients. Results of the Harvard Medical Practice Study II.” New England

Journal of Medicine. Vol. 324(6):377-84.

Lee, Douglas, Jack Tu, Alice Chong, and David Alter. 2008. “Patient Satisfaction and Its

Relationship with Quality and Outcomes of Care After Acute Myocardial Infarction.”

Circulation. Vol. 118:1938-45.

Lee, Fred. 2004. If Disney Ran Your Hospital: 9 ½ Things You Would Do Differently. Second

River Healthcare.

Levin, Henry, and Clive Belfield. 2003. “The Marketplace in Education.” Review of Research in

Education. Vol. 27:183-219.

Light, Donald. 1979. “Uncertainty and Control in Professional Training.” Journal of Health and

Social Behavior. Vol 20:310-22.

Localio, Russel, Ann Lawthers, Troyen Brennan, et al. 1991. “Relation Between Malpractice

Claims and Adverse Events Due to Negligence III . Results of the Harvard Medical

Practice Study.” New England Journal of Medicine. Vol. 325(4):245-51.

Lynn, Freda, Joel Podolny and Lin Tao. (2009). “A Sociological (De)Construction of the

Relationship between Status and Quality.” American Journal of Sociology. Vol. 115:

755–804.

Mazor, Kathleen, Steven Simon, and Jerry Gurwitz. 2004. “Communicating with Patients About

Medical Errors: A Review of the Literature.” Archives of Internal Medicine. Vol.

164(15):1690-97.

47

McGlynn, EA, SM Asch, J Adams, J Keesey, J Hicks, A DeCristofaro, and EA Kerr. 2003. “The

Quality of Health Care Delivered to Adults in the United States.” New England Journal

of Medicine. Vol. 348(26): 2635–45.

Paolino, Philip. 2001. “Maximum Likelihood Estimation of Models with Beta-Distributed

Dependent Variables.” Political Analysis. Vol. 9(4):325-346.

Papke, Leslie, and Jeffrey Wooldridge. 1996. “Econometric Methods for Fractional Response

Variables with an Application to 401(K) Plan Participation Rates.” Journal of Applied

Econometrics. Vol. 11(6):619-32.

Parsons, Talcott. 1951. The Social System. “Chapter 10: Social Structure and Dynamic Process:

The Case of Modern Medical Practice.” Glencoe, Ill., Free Press.

Podolny, Joel. 1993. “A Status-Based Model of Market Competition.” American Journal of

Sociology. Vol. 95(4):829-72.

RAND. 2006. The First National Report Card on Quality of Health Care in America. Santa

Monica, California.

Reinhardt, Uwe. 2006. “The Pricing of U.S. Hospital Services: Chaos Behind A Veil Of

Secrecy.” Health Affairs. Vol. 25(1):57-69

Retchin, Sheldon. 2007. "Overcoming Information Asymmetry in Consumer-directed Health

Plans." American Journal of Managed Care. Vol. 13(4):173-176.

Rodriguez, Hector, Ira Wilson, Bruce Landon, Peter Marsden, and Paul Cleary. 2007.

“Voluntary Physician Switching by Human Immunodeficiency Virus-infected

Individuals: a national study of patient, physician, and organizational factors.” Medical

Care. Vol. 45(3):189-98.

48

Scandlen, Greg. 2005. “Consumer-Driven Health Care: Just a Tweak or a Revolution?” Health

Affairs. Vol. 24(6):1554-58.

Schempf AH, Minkovitz CS, Strobino DM, and Guyer B. 2007. “Parental Satisfaction With

Early Pediatric Care and Immunization of Young Children: the Mediating Role of Age-

Appropriate Well-Child Care Utilization.” Archives of Pediatric Adolescent Medicine.

Vol. 161(1): 50-6.

Smithson, Michael and Jay Verkuilen. 2006. “A Better Lemon Squeezer? Maximum-Likelihood

Regression With Beta-Distributed Dependent Variables.” Psychological Methods

11(1):54-71.

Spence, Michael. 1974. Market Signaling: Informational Transfer in Hiring and Related

Screening Processes. Cambridge: Harvard University Press.

Studdert, David, Eric Thomas, Helen Burstin, Brett Zbar, John Orav, and Troyn Brennan. 2000.

“Negligent Care and Malpractice Claiming Behavior in Utah and Colorado.” Medical

Care. Vol. 38(3):250-60.

Studdert, David, Michelle Mello, Atul Gawande, Troyen Brennen, and Claire Wang. 2007.

“Disclosure of Medical Injury to Patients: An Improbable Risk Management Strategy.”

Health Affairs. Vol. 26(1):215-26.

de Vries, EN, MA Ramrattan, SM Smorenburg, DJ Gouma, and MA Boermeester. 2008. “The

Incidence and Nature of In-Hospital Adverse Events: A Systematic Review.” Quality and

Safety in Health Care. Vol. 17:216-23.

Wong HS, Zhan C, and Mutter R. 2005. “Do Different Measures of Hospital Competition Matter

in Empirical Investigations of Hospital Behavior?” Review of Industrial Organization.

Vol. 26:61-87.

49

Wu AW, S Folkman, SJ McPhee, and B Lo. 1991. "Do House Officers Learn from their

Mistakes?" Journal of the American Medical Association. Vol. 265(16):2089-94.

50

Appendix 1. Technical Medical Quality Indicators (Process of Care Quality Measures)

Heart Attack

Aspirin at Arrival

Aspirin Prescribed at Discharge

ACEI or ARB for LVSD

Adult Smoking Cessation Advice/Counseling

Beta Blocker Prescribed at Discharge

Beta Blocker at Arrival

Median Time to Fibrinolysis

Fibrinolytic Therapy Received Within 30 Minutes of Hospital Arrival

Median Time to Primary PCI

Primary PCI Received Within 90 Minutes of Hospital Arrival

Heart Failure

Evaluation of LVS Function

ACEI or ARB for LVSD

Adult Smoking Cessation Advice/Counseling

Discharge Instructions

Pneumonia

Oxygenation Assessment

Pneumococcal Vaccination

Blood Cultures Performed Within 24 Hours Prior to or 24 Hours After Hospital Arrival for Patients Who

Were Transferred or Admitted to the ICU Within 24 Hours of Hospital Arrival

Blood Cultures Performed in the Emergency Department Prior to Initial Antibiotic Received in Hospital

Adult Smoking Cessation Advice/Counseling

Antibiotic Timing (Median)

Surgical Care

Prophylactic Antibiotic Received Within One Hour Prior to Surgical Incision

Prophylactic Antibiotic Selection for Surgical Patients

Prophylactic Antibiotics Discontinued Within 24 Hours After Surgery End Time

Surgery Patients with Appropriate Hair Removal