Skip Navigation Archive: U.S. Department of Health and Human Services U.S. Department of Health and Human Services
Archive: Agency for Healthcare Research Quality
Archival print banner

This information is for reference purposes only. It was current when produced and may now be outdated. Archive material is no longer maintained, and some links may not work. Persons with disabilities having difficulty accessing this information should contact us at: Let us know the nature of the problem, the Web address of what you want, and your contact information.

Please go to for current information.

Section 2

An Agenda for Research in Ambulatory Patient Safety

Synthesis of a multidisciplinary conference to develop an agenda for research in ambulatory patient safety.
Recommendation 1: AHRQ should support a feasibility study to:
  1. Identify appropriate methods to conduct a large-scale study of the epidemiology of safety/error in ambulatory care.
  2. Perform pilot testing of the methodologies.
  3. Estimate the projected value and costs of such a study.15

While there have been no comprehensive studies of this magnitude, there is growing evidence from a variety of focused studies of risk, error, and injury in ambulatory care, as well as quality improvement approaches aimed at improving safety. Additional focused studies would generate information that could fill many of the gaps in our knowledge of the epidemiology of ambulatory patient safety.

Risks, Errors, and Injuries in Selected Areas

Though there have been few large scale research initiatives in ambulatory safety, there have been focused studies on high-risk ambulatory care sites, procedures, processes, and populations (Institute of Medicine 1999, Weingart et al. 2000). Studies have focused on particular components of ambulatory care (e.g., ambulatory surgery centers, office based surgery), population groups (e.g., the elderly), procedures (e.g., endoscopy), and there are a number of studies of medication use. These studies provide windows into ambulatory patient safety and make it possible to begin to assemble a mosaic of its scope and structure. Well-integrated segments of the health system, such as a large group practice that is integrated with a health plan, potentially offer a more complete microcosm of ambulatory care. But studies of individual organizations, particularly if atypical, may not be representative and the results could not be easily generalized.

Some practice sites (e.g., small physician practices) and/or practice content (e.g., office surgery, cosmetic surgery) may be associated with increased risk, a higher rate of incidents, or greater average severity of consequences, and so may deserve higher priority for research. Similarly attractive are opportunities in ambulatory care to test interventions known to be effective elsewhere in the health care system.

Ambulatory Surgery and Other Invasive Procedures

A rapidly growing number of surgical and invasive procedures of increasing complexity have moved from hospitals to ambulatory surgery facilities and increasingly to physician's offices—presumably with a progressive decrease in the sophistication of the equipment and personnel. It is projected that 25 percent of all elective surgical procedures will be performed in physician offices by 2002 (Quattrone 2000). Several reports have indicated that physicians are performing surgical procedures outside their specialty and without proper surgical training. In addition, improperly installed and operated equipment and the lack of rescue and resuscitation equipment further increase risk in these settings. There is increasing attention to ambulatory surgery from regulatory bodies, and a number of States are moving to regulate office based surgery through accreditation of surgery and anesthesiology practices. Specialty societies support voluntary accreditation and certification and are developing guidelines for this purpose (Prager November 2000).

The literature offers interesting information on both the safety of procedures performed in ambulatory settings and their safety relative to that of inpatient care. For example, a survey of all board-certified members of the American Society of Plastic and Reconstructive Surgeons (ASPRS)—917 respondents of 1200—reported a liposuction fatality rate of about 20 per 100,000 procedures or about 1 in 5000 (Teimourain and Rogers 1989, Talmor and Liberman 1997, Barillo et al. 1998, Rao et al. 1995).

Several influences that may reduce risk in inpatient care are less prevalent in ambulatory care. Most ambulatory sites are subject to less regulation, less peer interaction, and have less well developed policies and procedures (e.g., credentialing and privileging) to determine the training and experience required to perform surgery and invasive procedures, or to manage sedation and anesthesia. There are fewer checks on individuals with inadequate training and experience who wish to perform a high risk procedure such as liposuction.16 There are fewer standards and less regulation in outpatient settings, allowing greater variation in how well equipment is maintained (for example, monitoring devices), and in the education, training, and experience of personnel who use the equipment.

As volume continues to increase in office based surgery, complication rates may increase. This may be related to the shift of patients with greater co-morbidity to the outpatient setting. However, it is not clear whether the risks of procedures, such as liposuction, are greater in physicians' offices than in hospitals. For example, malpractice data from the Physician's Insurance Association of America (PIAA) from 1995-97 show that hospital-performed liposuction yielded three times the total number of malpractice settlements than from office based procedures (Coleman et al. 1999). The authors attributed this difference principally to differences in volume and local anesthesia techniques.

There is additional literature addressing the relative efficacy and in some cases safety of individual procedures that are increasingly being performed in ambulatory settings. For example, a meta-analysis focusing on outpatient cholecystectomy of 98 studies of 78,747 patients compared with 28 studies of 12,973 patients showed that laparoscopic cholecystectomy compared with open cholecystectomy had a higher rate of bile duct injury but lower mortality and appeared to be at least as safe (Shea et al. 1996). The laparoscopic procedure has allowed many more patients to undergo this operation (Escarce et al. 1995). However, later studies may indicate that differences are explained by differences in the type of patient and severity of illness over time. For this and many other procedures, there is relatively little information to identify the effects of setting and other factors on adverse events and injury.

Mezei and Chung, in two separate studies, have looked at factors contributing to return hospital visits, hospital readmissions and prolonged stay after ambulatory surgery performed at a major ambulatory surgery center in Toronto. They found that of 17,638 consecutive patients having ambulatory surgery, 193 readmissions occurred within thirty days (readmission rate of 1.1 percent). The complication rate was a very low 1.5 percent (Mezei and Chung 1999). However, in a study of 16,411 consecutive patients for factors leading to prolonged stays following ambulatory surgery, they found that the length of stay following ambulatory surgery is determined by type of surgery, type of anesthesia, and by any adverse effects of the procedure or the anesthesia (Chung and Mezei 1999). They predicted that elimination of such adverse effects would lower length of stay by 3.8 percent for patients receiving monitored anesthesia care and 9.6 percent for those receiving general anesthesia. However, the study did not assess preventability of or extent of injury from these events.

It should be noted that the movement of increasingly complex procedures from inpatient to ambulatory settings may offer opportunities to improve patient safety. A comparative study by Miller et al. (2000) of surgical site infections (SSI) in inpatient and ambulatory surgery settings found that the SSI rate in ambulatory surgery was only about 35 percent of that in inpatient settings and 38 percent of that of a "mixed" group. These differences are quite large, and were not explained by risk adjustment for severity of condition or by differences in patient populations. This raises the possibility that for some conditions and procedures—such as those that expose the patient to risks of surgical infections—the ambulatory setting may be safer.

Populations at Increased Risk

A large body of knowledge highlights differences in quantity and quality of health care that is delivered to underserved and minority populations, but there is little information on the whether or not the care provided is less safe than that provided to the general population (Institute of Medicine 1999, Burstin et al. 1992). A study of adverse drug events in a large random sample (93,950 elderly and 55,333 adult patients) in Quebec following a policy change that increased cost-sharing in the province's prescription drug benefit program showed that these increases resulted in reduced use of essential, appropriate drugs, a higher incidence of serious adverse drug events, and increased use of emergency services resulting from the adverse reactions (Tamblyn et al. 2001). The effects of cost-sharing differentially affected the poor and the elderly.

Some patient populations would be expected to be at greater risk in general and in ambulatory care in particular: persons with psychiatric disorders (including common mood disorder), comorbid illnesses, children and the elderly, the homeless, hearing impaired, language differences, those with culture differences, and those with reduced access for financial and other reasons. On both ends of the age spectrum, the very young and the elderly, observers are beginning to assess the impact of preventable error and the sources of error. Glauber et al. (2000), in an evaluation of a study by Baker et al. (1999) of the application of an effective guideline to identify high and low risk febrile infants for antibiotic treatment, found that even with an excellent tool to help make the correct decision, 13 of 18 emergency department physicians made errors in diagnosis and treatment. They suggested that the uniform distribution of error across physicians that they observed indicates a systemic problem. Decision points that led to abandoning the protocol were identified and system improvements suggested; for example, laboratory calculations of critical values should be provided along with the test results rather than leave those calculations to the physician. Errors in dosage can be very toxic to infants and children, and these errors are relatively common (Buck 1999).

On the other end of the spectrum, the elderly are particularly vulnerable to medical injury (Gurwitz et al. 1994, Gurwitz et al. 2000, Thomas and Brennan 2000, Rothschild and Leape 2000, Rothschild et al. 2000, Field et al. 2001). Those sixty-five years of age and over are more likely to have comorbid conditions and more severe illness. This increases risk and consequences of adverse drug interactions and other forms of medical and surgical error. In addition, the use of ambulatory care (as well as inpatient care) increases with age. The elderly make twice as many visits to medical providers as those under sixty-five (NCHS 1999a).

There is evidence that office based physicians not infrequently prescribe inappropriate medications for the elderly. A study using data from the 1992 National Ambulatory Medical Care Survey (NAMCS) found that about 7.75 million office visits by the elderly resulted in the prescribing of at least one medication from a list of 20 drugs judged highly inappropriate for the elderly and about 720,000 visits resulted in the prescribing of two inappropriate medications (Aparasu and Fliginger 1997). Overall, about 7.8 percent of the elderly who received prescriptions were given inappropriate medication, an improvement over an earlier study based on 1987 NAMCS survey data found that 23.5 percent of the elderly (6.5 million persons) were prescribed at least one drug from the same list of 20 highly inappropriate medications (Wilcox et al. 1994). Analysis of the 1996 Medical Expenditure Panel Survey yielded a similar picture (Zhan et al. 2001, Avorn 2001). Adverse drug events appear to be common in older outpatients who are taking multiple medications (Schmader et al. 1994, Hanlon et al. 1998). In their review of medical injury in older patients, Rothschild and Leape (2000) note that the shift from inpatient to outpatient care induced by managed care has made documentation of adverse events even more difficult. And, ambulatory settings are subject to the same systemic failures that lead to injuries in hospitals.

Adverse Medication Events

Adverse reactions from medication occur throughout patient care, and a growing number of studies have examined medication errors and adverse events in outpatients (Hanlon et al. 1998, Gandhi et al. 2000a, Gandhi et al. 2001a, Rothschild et al. 2001, Honigman et al. 2001a, Honigman et al. 2001b). Gandhi et al. (2001a) studied a random sample of 2,248 patients aged 20 to 75 reporting prescription drug use was drawn from 11 Boston area ambulatory clinics. Eighteen percent (18 percent) of these patients reported a drug complication. A medical record review reported complications in only 3 percent of the sample. Drug complications in these settings were common and resulted in increased use of medical care to manage adverse effects. Communication to patients of medication instructions and potential side effects was often inadequate, particularly when the language of the patient was not English. Because adverse drug events were often not documented in the medical record, it is likely that estimates of drug related complications in ambulatory settings are substantially under-reported.

Return to Contents

Sources of Data for Research on Ambulatory Safety

Though appropriately judged to be inadequate to create a comprehensive understanding of ambulatory patient safety, there are several existing sources of data and experience in health care that can be used in research.

Experience from Risk Management

In the absence of a large body of scientifically rigorous studies, a practical strategy may be to exploit existing data and information from the area of risk management, particularly the experience of liability insurance companies, and from the case study experience of individual health care delivery organizations that have internalized functions of quality assurance and risk management. Information from such studies could be used to sketch an outline of the epidemiology of ambulatory patient risk and injury. Data and experience from investigations of medical liability claims and from risk management in provider organizations can also provide detailed information on selected kinds of incidents. The investigations that physician liability insurers conduct into the factors leading to occurrences and incidents should provide a rich source of information about connections in ambulatory care and the high-risk aspects of care. These elements can complement large epidemiological studies and smaller studies focused on important aspects of ambulatory care.

Risk Management Activities of Liability Insurers

The most comprehensive datasets and some of the most well thought-out studies of ambulatory patient care safety come from malpractice liability and risk management companies. Liability insurers receive and investigate claims of injuries from adverse events and alleged errors. Some also receive notification of occurrences that an insured physician or organization identifies as potentially resulting in a claim. What can we learn about the types and patterns of injuries, adverse events, and potential errors from these occurrences? For example, investigations of incidents, which begin with the event and a search for factors leading to the event, provide a rich source of understanding how injuries occur. Each claim (and sometimes a potential claim) leads to "upstream" investigations that follow multiple leads, including the contribution of support processes, employing root cause analysis. Following these leads into care processes and support systems that are involved in the patient's care may expose "latent risks" in the system.17 The investigation may be followed by focus groups of providers to further understand the risks and sequence of events that resulted in the incident.18

For example, one study that investigated the question of preventable medical error in primary care ambulatory settings used a risk management database to isolate the prevalence of adverse events and assess whether or not an event was due to medical error and could have been prevented (Fischer et al. 1997). The investigators found that over a five and a half-year period that adverse events occurred at a rate of 3.7 per 100,000 clinic visits. Eighty-three percent (83 percent) were judged due to medical error and to be preventable. Fourteen percent of the errors resulted in permanent disability and 3 percent resulted in death. The authors concluded that serious adverse events appear to occur relatively infrequently in primary care but that the consequences can be very significant.

The Medical Group Management Association's Center for Research has worked with a number of malpractice liability firms and organizations. Several of these firms have supplied substantial data, information, and expertise to provide background information and insight into question of patient safety in ambulatory care: the COPIC Insurance Company, PIAA (Physician's Insurance Association of America, the St. Paul Insurance Companies, and the Harvard Medical Institutions' Risk Management Foundation (RMF) and their captive liability carrier, the Controlled Risk Insurance Company (CRICO).

COPIC, for example has had a "Participatory Risk Management" program in place since 1984. This program offers both positive and negative incentives for providers to adopt and use COPIC guidelines and use specifically designed tools and programs to improve practice patterns. The analysis of the COPIC claims database is quite helpful in understanding the occurrence of adverse events and their relationship to malpractice claims. Diagnosis specific claims for different specialties provide useful information on the distribution of adverse events across providers and in some cases, practice settings. COPIC has also developed a "Comprehensive Breast Management System" and worked with the MGMA Center for Research to evaluate its effects on the continuity and quality of care provided. COPIC has developed a taxonomy of adverse events and near misses, and COPIC's occurrence report system provides that basis for identifying errors and evaluating their effects (Harbrecht et al. 2000).

The Risk Management Foundation of the Harvard Medical Institutions and CRICO, its associated liability insurance company, have a long history of working to manage risk by defining areas of risk, identifying problems and designing programs to improve patient care. In addition to a comprehensive ambulatory claims data base, RMF has initiatives in an adverse events reporting system; best practices information from the "Office Practice Evaluation Program"; and numerous special projects aimed at breast cancer diagnosis and ambulatory care prescription and chemotherapy medications. CRICO supported a three year multi-site study, the Ambulatory Medicine Quality Improvement Project, that addressed many of these issues (Gandhi et al. 2000a, Haas et al. 2000).

The St. Paul Insurance Company is also involved in building comprehensive risk management databases and designing clinical risk management systems, and the data is derived from these processes is quite valuable. The St. Paul recently reported fourteen years of data and the effects of compliance with their Clinical Risk Modification Guidelines. These data indicate that risk management approaches have the ability of demonstrably improving patient care outcomes, identifying areas of emerging risk and reducing costs. This longitudinal data provides a foundation to evaluate patient safety issues in the light of changing patterns of delivery, changes in patient populations and demographics and changes in financing and reimbursement. All have an impact on the shape of the health care delivery system and impact on the safety of the population it serves (MMI Companies 2000).

The aggregated experience of liability insurers' investigations can provide a starting point for constructing an epidemiological picture of ambulatory patient safety. For instance, several liability insurers have found that the largest single category of errors leading to claims and awards are delays and failures to diagnose (e.g., breast cancer), and the diagnostic process occurs largely in ambulatory care. It may be possible to begin with the database of liability claims that is generated by a known population and use one or more methods to identify the larger set of occurrences for that population. For example, this approach might be applicable to self-contained integrated deliver systems with captive liability insurers (i.e., "closed" systems) such as Kaiser Permanente, or the institutions that are associated with the Risk Management Foundation of the Harvard Medical Institutions and CRICO.

However, the pattern of liability claims or settled cases would not be expected to provide a full and accurate picture of errors or adverse events, so it would be important to understand how the rates and patterns represented in these data would differ from those found in a full and accurate epidemiological study. Events that lead to actual or potential claims are thought to be the "tip of the iceberg" and do not represent the whole spectrum of incidents and near misses. Databases of adverse events developed from risk management activities probably underestimate the prevalence of adverse events, particularly of less severe injuries that are not likely to generate a malpractice claim. Nevertheless, in the absence of systems specifically designed and implemented to identify and track adverse events and their sources (Fischer et al. 1997), these data are valuable. And liability companies have a wealth of experience in dealing with the consequences of medical injury and error (Bogardus et al. 1999, Bovbjerg 1992, Taragin et al. 1992). Liability insurers not only investigate incidents and occurrences, they also develop interventions to reduce the rates of incidents and their consequences, and there is much we can learn from their experience.

Research Questions: How can we use investigations of liability claims, sentinel events, and other incidents to lay out the epidemiology of ambulatory patient safety? Are there methods to draw conclusions about the larger universe from these incident data? Also, how can we increase the number from ambulatory care? To what extent (and with what data) can we make inferences about larger patterns of events, and about populations? How can they be used in the construction of a conceptual model of ambulatory care and its risks? How can this information be used to increase awareness of persons in health care and the public about the kinds of risks, the factors that increase risk, and the steps that could be taken to reduce risk/increase patient safety in ambulatory care? How might the methods used in identifying and investigating these events be changed to improve the usefulness of these data for research?

Recommendation 2: Support research that examines and evaluates claim and incident data from liability insurers, and how the rates and patterns of incidents and injuries from these sources would be expected to differ from the "true" rates and patterns. Specific efforts could assess the value of case studies. Support evaluation of interventions by liability insurers that are designed to improve ambulatory patient safety, and identify those that are effective and should be widely disseminated.

Provider System Experiences in Risk Management

The collective experience of providers in general, and integrated systems in particular, should provide a relatively rich source of data that is particularly well suited to the development of hypotheses and design of rigorous studies. Risk assessment and risk management activities designed to gather and analyze data in order to address specific problem areas have generated a substantial body of case-study information that yields valuable information on the extent of individual patient safety problems and on their solution. For example, a large multi-specialty group practice and integrated delivery system located in the Pacific Northwest has generated methods of identifying, categorizing and addressing issues and potential problems in patient safety in the ambulatory care delivery systems for which they are responsible and accountable. The discussion below draws in part on that system's experience.

Risk management programs frequently break errors down into two categories for investigation and intervention: those of omission and those of commission (Kravitz et al. 1997). One large integrated system classifies high risk areas for medical error in ambulatory care settings as:

  1. Failure to diagnose.
  2. Screening and follow-up, which are errors of omission.
  3. Issues of patient identification.
  4. Complex technology and less well trained personnel.
  5. Oversedation, which are errors of commission.

Documentation of these investigations and experience with interventions have established institutional bodies of knowledge that represent a largely untapped source of data on patient safety in ambulatory settings. For example:

  • Failure to diagnose may be the result of diagnostic system design. For example, in the ambulatory setting the provider might order a number of laboratory and radiology tests and an electrocardiogram following a routine physical examination. The test results may come back to the provider's office on different days and on different pieces of paper. If the provider performs five examinations that result in such tests in one day, this will generate around fifty separate pieces of paper that must be managed and assimilated. The provider is most often accountable for notifying each patient of any abnormal or suspicious test result. There is great risk that one or more of these pieces of paper will be misplaced or never received, with the potential for an error with serious consequences. One of the most common is that women may not be notified of a suspicious mammogram that needs follow-up. Such system failures are a "root-cause" of malpractice claims in this area. There are many potential approaches to this situation. Three common approaches have been:
    • Assign a full time non-professional employee to batch, manage and audit the paper slips.
    • Assign an office R.N. to review each slip and contact patients for all normal results and leaving the abnormal result contact to the provider.
    • Schedule each patient for a follow-up examination within a two week period.
    There are varied costs and benefits to each approach. Other innovative approaches include systems that enable the patient to obtain results at any time over the telephone or over a secure Internet Web site.
  • Providers are increasingly being held accountable for reminding patients to schedule screening and follow-up studies at specific age intervals. Tort case law appears to be reinforcing this responsibility. Mammograms for women fifty and over is an example. Follow-up of identified disease-specific problems, particularly diabetes and hypertension is another area of contention. In ambulatory care there are usually no systems in place to tell the provider when they last saw patients with specific diseases.
  • Integrated systems have begun to institute identification and reminder systems for both screening and follow-up care. For example, one screening system interacts directly with the patient with minimum involvement with the provider's office. In the case of mammograms, women overdue for one are notified on each quarter. It is estimated that this notification system, in its first three years, resulted in an additional 800 women being screened and identification of 5-6 cases of cancer that would have gone undetected, thus preventing errors of omission. Disease-specific management systems are intended to ensure that patients with a particular condition receive needed diagnostic and therapeutic services by working with patients and their caregivers (Wagner et al. 1999).
  • It is surprisingly common in ambulatory care for patients to be misidentified. Patients are not commonly given bracelets or cards. This produces opportunity for medication, treatment and/or education error. In a fast paced office environment with many tasks and many pieces of paper being completed at the same time there are many opportunities for such error. In some cases patients will simply not understand that they are getting the wrong treatment. One integrated system has found these kinds of errors to be substantial, and is implementing a system to manage loose paper notes and test results, with checks of patients' names.
  • Complex and seriously ill patients are often managed in ambulatory care by less highly trained personnel than in a hospital, and they are usually supported by a less extensive infrastructure. There appear to be significant problems with triage systems and telephone consultations being staffed with personnel who have in general a lower skill-set than those who provide these services in inpatient settings. Use and monitoring of equipment, such as microwave therapy machines in urologist offices, should be carried out by trained and certified personnel. There are many documented cases of burns occurring because of improper use and monitoring by lower skilled employees. This situation is more likely to occur in smaller practice settings but may occur in any setting. The real issue is that of cost-effectiveness and matching the optimum level of care with the appropriate level of skill and resultant cost. In an era of increasing competition, the issue is significant and growing.
  • Over-sedation as an issue tied to higher technology in settings where there is considerable variation in the ability to manage safely—one of the most significant medication issues in this and other delivery systems. A team approach and certification of both physicians and registered nurses to validate required skills are required.
  • Another large integrated system with a captive insurer collects risk management information at three levels:
    • Incident reports, which include the in-depth investigation of adverse events and claims.
    • Reports of "potentially compensable events" (PCEs) that are essentially "near misses" and other incidents that reveal risk but may not generate claims.
    • Traditional morbidity and mortality and adverse outcome conferences to examine selected deaths and injuries.
    The captive insurer is a partner in these activities, and helps the organization to internalize the costs as well as benefits of this work. The organization also recognizes the implications of these risk management activities for improving safety and for credentialing.

There is wide variation among provider organizations in the awareness of specific issues, approach to solutions, and availability of resources to address identified problems. However, the documented experience of providers and delivery systems charged with dealing with identified problems provides a source of useful information to inform and guide more comprehensive and rigorously designed studies.

Return to Contents
Proceed to Next Section

Page last reviewed December 2001
Internet Citation: Section 2: An Agenda for Research in Ambulatory Patient Safety. December 2001. Agency for Healthcare Research and Quality, Rockville, MD.


The information on this page is archived and provided for reference purposes only.


AHRQ Advancing Excellence in Health Care