A High Performance Hospital

I interviewed Dr. Lucian Leape, adjunct professor of Health Policy at Harvard School of Public Health, who is a pioneer in hospital safety and quality. When I asked how he framed the difference between the two, Leape said that quality and safety are always found together. He thinks of quality as process measures; for instance, did the patient get antibiotics before surgery? He thinks of safety as the outcome measure of quality; for instance, did the patient have a wound infection?

According to Leape, two US hospitals are considered the best examples of safe, high quality hospitals, LDS (Latter Day Saints) Hospital in Salt Lake City, Utah, and the Mayo Clinic in Rochester, Minnesota. I flew to Salt Lake City to interview one of the leaders in the quality movement, Dr. Brent James.

James, who was a cancer surgeon before he became consumed with creating quality health care systems, came from Harvard to LDS Hospital in 1986. The hospital was an early adopter of information systems. In 1964, a visionary outfitted LDS hospital’s intensive care units with computers that filled seven office-sized rooms. LDS Hospital was part of Intermountain Healthcare, a statewide group of hospitals and clinics.

At LDS, James met Steve Busboom, Intermountain’s Vice President of Finance, who was struggling to find a system to bill insurers more accurately. Hospitals all have a list of billable services called a charge master and Busboom was struggling with 25,000 items on the LDS charge master. James and Busboom joined forces to examine not only the variations in care but also the cost of these variations in care. I asked who paid for the work and Dr. James said, “Steve talked the system into it. He was working on billing so they thought they might save some money.”

James had studied the work of John Wennberg and his colleagues at Dartmouth Institute for Health Policy and Clinical Practice. It was this group’s work that showed the 160 percent regional variation in Medicare spending and demonstrated that high-cost areas were also low-quality areas, with increased mortality rates. James expected to find variations among practices within the same hospital. His first study was to examine six common procedures. James said, “I had no appreciation of how common, common procedures are.” A small number of procedures and diagnoses consume a vast amount of hospital resources, just as has been observed in the health care system as a whole where ten percent of patients consume seventy percent of resources.

James illustrated his process with a procedure called a transurethral prostatectomy (TURP). A TURP is a common procedure performed in elderly men whose prostate glands enlarge and obstruct the urinary tract, preventing or slowing urination. In performing TURPs, urologists insert a narrow tube up the penis and then remove a portion of the prostate gland in order to increase urination, boring an opening for the urine to pass through the prostate gland. The investigators analyzed the variations in care among TURP patients and found ninety separate factors. Although the patient selection and the way the procedure is performed is about as standardized as any intervention in medicine. James found ninety different variables that could be used one or more times in the care of a patient undergoing a TURP. He said that the closest the physicians came to the use of the same variables was a sixty percent variation from high to low (that is, if one physician was using ten of something to get the job done the closest any other physician came was to use sixteen of the same thing to get the same outcome on the same patient). The widest range of variation among physicians doing a TURP, within the ninety factors tracked, was a 460 percent variation (ten versus fifty-six of the same thing to get the same outcome).The average pattern for any two patients was a 160 percent variation among the ninety factors. To emphasize his point he said, “I could scarcely figure out what a TURP was.” They also found a one hundred percent difference in the cost of care from the high to the low end.

James said he always thought that some surgeons were just naturally better than others, a view I also held. But what they found at LDS was that in no area of management was one surgeon always the best. The length of procedures varied from thirty-eight minutes to ninety minutes, and the faster the surgeon, the more prostate tissue was removed--varying from thirteen grams of prostate for slow surgeons to forty-two grams for fast surgeons. This finding was counterintuitive. The patients of the slower surgeons also had postoperative obstruction of the urinary tract more frequently, a complication which requires additional interventions, longer stays, and thus more cost.

This data was presented to the urologists, who could identify themselves but not other surgeons. James’s group did not tell the urologists what to do with the data. James said he was sure that the slowest surgeon would never talk to him again, but some time later the slowest urologist called him and thanked him. James said that the doctor’s comment was, “I talked with my colleagues and they could tell where they were in the prostate gland by closely observing the nature of the prostate tissue they were resecting. This means they could remove tissue very quickly at first, and then slow down when they got close to potentially dangerous margins – areas where the removal of tissue could damage the prostate. I had been picking my way more slowly and actually could not remove as much tissue safely. I adopted their technique.” The reason why the slow surgeons’ patients’ urinary tracts developed post-operative obstructions was because the slow surgeons cut too close to the outflow of the bladder, causing scarring that obstructed urine.

One year later when the data from the same surgeons was examined, the slow surgeons were faster, removed more tissue, and their patients had no more complications than the others. The range of variation among the ninety factors was less, complications were fewer, and cost was reduced. The surgeons learned from one another. As James’s process unfolded, culture in the hospital began to change service by service.

The Intermountain Healthcare administration got behind James’s work when the method showed them how to reduce the cost of hip replacements from $12,000 to $8,000. At about the same time, another group of investigators at the hospital showed that wound infections were increased by 240 percent if surgical patients received their antibiotics after surgery rather than before it. The hospital made administration of pre-surgery antibiotics standard practice, saving large sums and reducing infections.(1) In 1992 the scope of James’s operation changed from a research unit within the hospital to a management tool for the hospital. Money was being saved, and patient outcomes improved in the bargain.

As our interview progressed, Dr. James commented on how change is adopted in hospitals. In 1987 he attended a four-day conference taught by W. Edwards Deming where he began to ponder how to convert Deming’s concepts to medicine. Deming was the mind behind the manufacturing and business transformation of Japan in the years after World War II. His great contribution was to answer the question in a Japanese national experiment of whether the money and effort expended to reduce variation in manufacturing processes saved money. The answer was yes; it earned Japanese manufacturers a lot of money because the value of their products improved so much. At one time the variance in measurements of a Japanese engine part was half that of a corresponding American engine part. The manufacturing processes that led to the Toyota and Lexus cars, which consistently earn among the highest quality ratings, stemmed directly from the use of manufacturing processes championed by Deming—processes that reduced variability.

James told me that Deming found that if the behaviors of the square root of the number of people in any group changed, then the behaviors of the whole group would change. That is, to change the behaviors of a group of nine people, three had to be turned. James said, “It works better in medicine if it is the right three people.” He emphasized that no doctor’s behavior could be changed by force for very long; the key to changing the behavior of physicians in the long term was to give them information. One of James’s experiences provides an example of this.

Delivery of babies is the most common procedure performed at Intermountain Healthcare’s hospitals, so James’s research group turned their focus to deliveries. Induction of labor is a procedure in which a hormone is administered that causes a pregnant woman’s uterus to contract, bringing on labor and causing delivery. The most common reason for induction is when medical conditions or an overdue fetus make immediate delivery safer than waiting for natural labor to begin. The American College of Obstetricians and Gynecologists recommends induction of labor only when the baby is fully developed (at least thirty-nine weeks) and when the woman’s cervix is ready. Induction is not benign because it leads to hard labor that can increase the chance of injury to the fetus.

The use of induction more than doubled from nine percent of deliveries in 1989 to 20.5 percent in 2001.(2) In upstate New York in 2001, twenty-five percent of inductions had no apparent medical basis, and the induction rate of laboring women varied from ten to thirty-nine percent among hospitals.(3) Like many things in medicine, a number of factors enter into the decision to induce; some of those factors are not medical—the convenience to the doctor and the mother of having a pre-scheduled delivery rather than waiting for labor to begin spontaneously. The variation in induction rates should not be so great if the procedure is being done for strictly medical reasons--the same thesis as Wennberg’s regional variation data but played out in one hospital.

James’s group found astonishing variation in the rates of induction among practice groups, among hospitals in the Intermountain system, and among physicians within groups. They presented their data to the system’s obstetricians. The data showed that among the Intermountain hospitals, 28 percent of elective inductions were inappropriate by the American College of Obstetricians and Gynecologist guidelines and that induction was associated with a higher rates of admissions to the neonatal intensive care unit..(4) The obstetricians were asked to change guidelines as they saw fit, but to reach an agreement. The consensus among the obstetricians was to decrease the percentage of inductions from twenty-seven percent of births to five percent. Most of that change was accomplished within a few months. Three years later, the rate had dropped to 3 percent. James told me that he simply kept providing the obstetricians the data. He noted that one group held out, never altering their procedures, but after a few years, peer pressure forced even this recalcitrant group to conform. James said that this is how it works. Five percent of doctors buy in at first, then more, and the last five percent are eventually brought in by peer pressure.

James emphasized that neither he nor anyone else ever told a doctor what to do. When a standard protocol was developed, it was disseminated among participating doctors and they were asked to change the protocol as they saw fit, not to approve or disapprove it. The lesson that was learned over and over is that best management arises when the ones who deliver care design the protocol. He viewed his function as providing the framework and the information needed to standardize management and then to report regularly on compliance with that management. James believes that for efficiency in medical practice to become permanent and not just the result of eternal vigilance that payment policy will have to be aligned with quality practice.

It Is About Culture, Not Computers

After talking with Brent James, I walked up the hill from the Intermountain corporate offices to LDS Hospital to interview Dr. Terry Clemmer, professor of internal medicine at the University of Utah and director of critical care medicine at LDS Hospital. Clemmer is not only a health care researcher but also a frontline physician who cares for the sickest patients in the hospital, those in the ICUs. We talked in his small windowless office next to one of the intensive care units (ICUs). I was curious to hear his views on Intermountain’s quality and information technology process because he has to live with it as he takes care of patients. Clemmer told me that by the 1980s the hospital already had a sophisticated information system, and coincident with James’s arrival, his critical care group had decided to move into protocols of management of ventilated patients. Prior to that each doctor managed ventilated patients as he or she saw fit, the normal way of practicing medicine.

They assigned ventilated patients to two groups. One group was managed by the doctors but with the aid of a computerized physician support tool. The support tool displayed information on computer screens at the bedside, advising the managing physician--for example, if A exists then do B. These computerized tools are experimental and have generally not improved management, but they do necessitate developing a standardized protocol for managing a particular kind of patient.

Clemmer’s group tested such a protocol and he said, “We decreased the mortality rate in both groups of patients by four times, and we concluded that management protocols improve care whether a computer is used to remind doctors or not.” I have heard such a story before. It seems to be a common lesson that computers in medicine are better at measuring and recording processes and providing feedback than at aiding physicians in their decision making.

Clemmer’s group learned other lessons. Clemmer and James both told me that the development of protocols has to be done at the level of those who manage the patients, but each meant something different. James is an administrator and a researcher; he meant the doctors managing the patients had to design processes, not administrators. Terry Clemmer meant the nurses, pharmacists, and respiratory therapists at the bedside, not necessarily the doctors. Clemmer told how his group of physicians developed a protocol for managing patients on ventilators for lung failure. When the doctors presented the protocol to the respiratory therapists, they all laughed. He said, “We changed the protocol because they were right and also because they are at the bedside and they had to buy in to it.”

Clemmer gave another example of where good protocols come from. High blood glucose levels cause all manner of complications and are very common in ICU patients. Insulin is the treatment, but use of insulin is tricky because patients vary in their response to it, and lowering the glucose too much can cause insulin shock which can be fatal. High glucose exerts its bad effects over a period of time, so the understandable tendency in hospitals is to err on the side of too high rather than too low. Clemmer credited managing blood glucose in a lower range with reducing the mortality rate in the hospital’s ICUs. Such a protocol does not require more effort by the doctors but does require more attention by the bedside nurses who must frequently check the glucose values, alter the insulin dose, recheck the blood glucose, and then instantly respond if the insulin overshoots and drops glucose to dangerous levels.

Clemmer turned to his computer screen as we talked and pulled up the blood glucose profiles for the last twenty-four hours of every patient in the LDS Hospital ICUs. At a glance he could tell if the protocol was being followed. Clemmer summarized his view of hospital quality, “It’s a matter of being consistent, reliable, and doing what you say you are going to do.”

I asked if the hospital saved money on its quality efforts. Clemmer said that LDS had reduced the cost of ICU patients by $2 million a month by reducing complications and mortality. “But we do not get to keep the money,” he commented. The explanation for this is that Medicare pays hospitals by the discharge diagnosis and most insurers simply pay by the number of interventions a patient is subject to. Hospitals get paid more if, upon discharge, a patient has had a lot of complications. For instance a hospital is paid a certain amount for a patient with heart disease admitted for a coronary artery bypass. If the patient is discharged with no complications, the hospital is paid less than if the patient had a myriad of complications. More work earns more pay from Medicare and much more from private insurers, even if sloppy medical care is the cause of the additional work.

This approach leaves hospitals and doctors without the slightest financial incentive to reduce complications and a powerful incentive to tacitly allow them. The payment system pits a basic tenet of medical practice “first, do no harm” against the self interest of doctors and hospitals.

LDS partially solved this problem by becoming an insurer so that at least for its own beneficiaries; good medicine is not unprofitable medicine. A hospital payment system that rewards complications, as the US system does, is perverse. The US is getting what it is paying for in every area, paying for quantity, not quality.

I asked Clemmer if work like at Intermountain is possible without fully functional electronic medical records. He said that computers were necessary but not sufficient. What is needed is computers with the will to use them--- an information system without a culture of safety and efficiency is wasted money. Dr. Clemmer advised me to look into a story that is now famous in quality circles, the Cedars-Sinai story.(5)

Cedars-Sinai, a stone’s throw from Beverly Hills, has the reputation of a cutting-edge hospital. It was not out of character that in 2002 the hospital invested $34 million on an information system and that it had already computerized several patient care areas. But when the process was extended throughout the hospital, the medical staff revolted. The system had been programmed not to accept a medication order from a doctor if the medication ordered interacted with one the patient was already taking. However, doctors frequently and deliberately combine drugs that interact, and the computer stopped them cold. Even in executing uncomplicated orders, the doctors found the number of questions they had to answer before the computer would accept their order slowed them down. Only a fraction of the two thousand doctors with privileges at the hospital were involved in the development of the information system. In 2002, within three months of installing the system, the hospital shelved it.

Until there is a business case for doctors and hospitals to use information technology, it will be an add-on, not an integral part of practice. Payment should be changed so that doctors and hospitals are paid for efficiency. Then there would be a business case for the use of information technology.

1. David C. Classen, R. Scott Evans, Stanley L. Pestotnik, et al., “The Timing of Prophylactic Administration of Antibiotics and the Risk of Surgical-Wound Infection,” New England Journal of Medicine 326, no. 5 (January 30, 1992): 281-286.
2. Joyce A. Martin, Brady E. Hamilton, Stephanie J. Ventura, et al., “Births: Final Data for 2001,” National Vital Statistics Reports 51, no. 2 (December 18, 2002).
3. J. Christopher Glantz, “Labor Induction Rate Variation in Upstate New York: What is the Difference?” Birth 30, no. 3 (September 2003): 168-174).
4. “Reducing Inappropriate Induction of Labor: Case Intermountain Health Care,” The Commonwealth Fund, November 29, 2004. Accessed online 2007: http://www.commonwealthfund.org/innovations/innovations_show.htm?doc_id=250148.
5. Ceci Connolly, “Cedars-Sinai Doctors Cling to Pen and Paper,” Washington Post. March 21, 2005.