Key Points

Reducing patient harm is complicated by problems inherent in accurately categorizing, detecting, and measuring harm. This paper summarizes the medical literature attempting to address such challenges, our institution’s approach to similar issues, and the results we have achieved by focusing efforts around a preventable harm index (PHI).

Categorizing harm: What harm is “preventable”?

A common view is that not all harm is preventable and separating inevitable harm from preventable harm is complex. Given current resources, efforts to reduce patient harm should first focus, where possible, on preventable harm events. Importantly, “preventability” can change over time. For example, VAP in pediatric patients was once considered the “cost of doing intensive care unit business.” It is now known that VAP can be effectively eliminated by carefully following a VAP prevention bundle. In the case of VAP, yesterday’s inevitable harm has become today’s preventable harm. This is likely true of other forms of today’s “inevitable harm.” Thus, defining what harm is preventable is an iterative process that requires continually challenging the notion of “preventability.”

A recent review by Nabhan and colleagues adds clarity about the types of harm that might be preventable and summarizes published report’s use of the term “preventable harm.” This summary describes 7 themes commonly found in a preventable harm definition; however, 3 are most compelling and common: (1) Presence of an identifiable modifiable cause (cited in 44% of references reviewed), (2) adoption of an intervention or process will prevent future occurrences (cited in 23%), and (3) lack of adherence to guidelines implies preventability (cited in 16%). Furthermore, 50% of the studies that outlined specific harm severity classifications used item 1 above as the operational definition for preventable harm. At Nationwide Children’s Hospital, we have adopted a definition for preventable harm that incorporates the principals noted. We believe this definition navigates the complexities between preventable and inevitable harm. Specifically, designation as preventable does not necessarily mean that each occurrence in question was preventable at the time, because this may be impossible to determine. Instead we have chosen to classify domains of harm as “preventable” as a call to action to improve patient care. This definition is designed to change the thinking that harm occurrences are expected complications and, to instead see them as challenges to reduce and ultimately eliminate. In some situations, when preventive policies were well known and not followed, or known complications were not adequately prepared for, such events will be adjudicated as preventable.

Categorizing harm: What harm is “preventable”?

A common view is that not all harm is preventable and separating inevitable harm from preventable harm is complex. Given current resources, efforts to reduce patient harm should first focus, where possible, on preventable harm events. Importantly, “preventability” can change over time. For example, VAP in pediatric patients was once considered the “cost of doing intensive care unit business.” It is now known that VAP can be effectively eliminated by carefully following a VAP prevention bundle. In the case of VAP, yesterday’s inevitable harm has become today’s preventable harm. This is likely true of other forms of today’s “inevitable harm.” Thus, defining what harm is preventable is an iterative process that requires continually challenging the notion of “preventability.”

A recent review by Nabhan and colleagues adds clarity about the types of harm that might be preventable and summarizes published report’s use of the term “preventable harm.” This summary describes 7 themes commonly found in a preventable harm definition; however, 3 are most compelling and common: (1) Presence of an identifiable modifiable cause (cited in 44% of references reviewed), (2) adoption of an intervention or process will prevent future occurrences (cited in 23%), and (3) lack of adherence to guidelines implies preventability (cited in 16%). Furthermore, 50% of the studies that outlined specific harm severity classifications used item 1 above as the operational definition for preventable harm. At Nationwide Children’s Hospital, we have adopted a definition for preventable harm that incorporates the principals noted. We believe this definition navigates the complexities between preventable and inevitable harm. Specifically, designation as preventable does not necessarily mean that each occurrence in question was preventable at the time, because this may be impossible to determine. Instead we have chosen to classify domains of harm as “preventable” as a call to action to improve patient care. This definition is designed to change the thinking that harm occurrences are expected complications and, to instead see them as challenges to reduce and ultimately eliminate. In some situations, when preventive policies were well known and not followed, or known complications were not adequately prepared for, such events will be adjudicated as preventable.

Measuring patient harm

Traditionally, studies examining improvements in patient safety and medical error prevention refer to the rate of harm in a given population. This approach is intended to allow inter-institutional comparison of event rates (benchmarking) and to normalize event frequency according to patient volume or other demographic characteristics. Rate determination offers many challenges, including defining both the number of harm events (the numerator) and the number vulnerable to that harm (the denominator).

Difficulties in determining numbers of harm events (numerator) include variation in event definitions and event detection methods. Little consistency exists in defining harm events. A recent review of 45 studies of adverse drug events (ADEs) yielded 26 different definitions. Consequently, prevalence of drug errors ranges from 2% to 75%. This issue is magnified when studies involve multiple institutions and/or multiple forms of harm, making data comparison between institutions difficult or impossible. Although consistently defining harm event to allow comparison of data is a daunting task, it becomes even more difficult when events such as “failure to diagnose” or near misses are included. Some harm domains are more reliably defined and applied. Specific examples include pressure ulcers (PUs) and hospital-acquired infections. Multi-institutional, regional, and national collaboratives are more often using the National Healthcare Safety Network definitions for hospital-acquired infections or the Wound, Ostomy and Continence Nurses Society Practice Guideline PU definitions.

Varying detection methodologies can yield different numbers of harm events. Each detection method has limitations. The most common detection method is voluntary reporting or occurrence reporting. Multiple studies document significant underreporting of events, ranging from 2% to 8% of total actual events. Retrospective or concurrent chart review is a second detection method frequently combined with occurrence reporting. This method is disadvantaged by inconsistency in terminology and in defining what actually constitutes an event, making it difficult to interpret the data. Trigger-based chart review is a third method that looks for specific “occurrences, prompts, or flags found on review of the medical record that ‘trigger’ further investigation to determine the presence or absence of an actual adverse event.” Although global trigger tools find considerably more events than other detection methods and are reasonably reliable, the process is labor intensive. Electronic automation may ultimately improve the usefulness of this methodology. Mandatory reporting of some specific harm events has also been suggested as a method to reduce underreporting of harm events; however, this approach may actually send reporting “underground” owing to fear of reprisal. For certain types of harm, such as hospital-acquired PUs, active surveillance using scheduled rounds can be useful to enhance reporting and detection. Importantly, results depend on the frequency of surveillance rounds (detection methods). For example, weekly “hospital-wide skin rounds” detect many more PUs than quarterly rounds conducted just in intensive care units.

Defining the population vulnerable to harm (denominator) is problematic. For example, for ADEs the denominator might be the total number of patients, the number of patient-days, the number of patients receiving that drug, the number of dispensed doses of all drugs, or the number of dispensed doses of that particular drug. Each denominator definition would yield a different rate for the same numerator. This same challenge exists for patient falls, PUs, and hospital-acquired infections.

To compare outcomes between institutions or programs, it is imperative that definitions and detection methods be consistent; otherwise, such comparisons are not value added. This issue will grow as hospital reimbursement and reputation are linked with harm event rates and as outcomes linked to individual practitioners become public (eg, Pennsylvania and New York). To date, public disclosure of harm and outcomes data has had minimal impact. As will be discussed subsequently, the PHI obviates some of these concerns.

Patient safety indicators: A practical measure of harm?

The Patient Safety Indicator (PSI) is a recently described composite measure of patient harm. The Agency for Healthcare Research and Quality developed a list of indicators, including patterns of diagnoses and/or procedures, that suggest that a potentially preventable adverse event may have occurred. The PSIs were developed through literature review, clinical input from physician groups, and empiric testing. This composite system can help hospitals to identify admissions or groups of admissions that might include populations at increased risk for preventable harm. Not surprisingly, hospitalizations that contain a PSI, as a group, have excess mortality, lengths of stay, and cost compared with those hospitalizations without a PSI. The PSIs do not lend themselves to tracking specific types of harm reduction over time and suffer from many of the problems associated with harm data derived from administrative sources.

The PHI: A different type of harm metric

In 2010, we described the use of a patient harm metric that does not involve a denominator: The PHI. This metric has several important features that make it particularly useful. First, the PHI sums (aggregates) individual events of harm and as such closely tracks numbers of patients harmed. Because the PHI uses only numbers of events, there is no denominator; therefore, it is not an event rate. This makes the number more personal (how many children are injured), more understandable, and more motivational. For example, a rate of 0.1 ADEs per 1000 dispensed doses seems reasonably low; however, at Nationwide Children’s hospital, it represents more than 125 patients per year who had an ADE. A focus on actual patients harmed rather than rates can inspire staff and leaders to make the changes necessary to reduce the numbers of such events.

At Nationwide Children’s hospital, preventable harm is tracked in 8 separate domains ( Table 1 ). All harm events are adjudicated as preventable using the harm definition previously described. Further, we have adopted internally consistent definitions for each harm domain based on national or regional publications where available. These domains of harm may not represent all hospital harm and other hospitals could choose to use additional domains as long as internal consistency of harm definitions and case identification is maintained. For all domains of harm, our event detection system includes voluntary event reporting, trigger tools, customer feedback including complaints and grievances, and active surveillance (PUs in particular). Total hospital-acquired infections are the sum of all hospital-acquired infections including central line associated blood stream infections, VAPs, catheter-associated urinary tract infections, and surgical site infections. ADEs for PHI purposes focus on severity levels 4 through 9 (D-I) as defined by the National Coordinating Council for Medication Error Reporting and Prevention. Our version of a rapid response team, the Assessment Consultation Team–preventable codes include any respiratory, cardiac, or cardiopulmonary arrest occurring outside the ICUs, the perioperative area, or the emergency department and are deemed preventable if detectable deterioration occurred before the event and an Assessment Consultation Team consult was not activated before the code. A multidisciplinary Code Blue Committee reviews and adjudicates each event. Postoperative complications include (1) unplanned returns to the operating room within 7 days of operation without death and (2) unanticipated death anytime during the hospital stay. Not all operative complications are preventable. To determine preventability, the following questions must be answered: (1) Is the event a known complication? and (2) Were all prevention steps followed? If yes answers these questions, then the event may not be preventable. Previously unknown complications are not considered preventable. Serious preventable falls are those that result in injury (eg, laceration, fracture) or require additional radiologic imaging (eg, simple x-ray or computed tomography). Developmental falls are excluded. PUs include grades 2 through 4 as defined by the Wound, Ostomy and Continence Nurses Society. Serious safety events (SSEs) are defined as variations from best or expected practice that reach the patient and are followed by serious harm up to and including death. “Serious harm” is defined as at least “moderate temporary harm” using specific criteria defined by our consultants, Healthcare Performance Improvement, LLC. SSEs are counted once. For, example if a surgical complication rises to the level of an SSE, it is listed in the tally for surgical complications and SSEs, but counts only once in the PHI total.

The count in each harm domain is the number of events (not a rate); therefore, disparate events can be summed to a total that reflects all events of patient harm for any given time period. As we began our safety journey that number was large, and served as a robust motivational call to action. We believe this would apply to most hospitals as they begin a safety campaign to reduce preventable harm.

The PHI is primarily intended as an internal, hospital-specific metric. As a result, harm definitions must be precise (if possible) and internally consistent within 1 institution or system of related hospitals. However, because the PHI is not intended for cross-hospital comparisons, it is not necessary to achieve consensus on definitions between hospitals. The PHI as an internally consistent, hospital-specific metric should be linked to a commitment by each hospital that uses it, to drive the PHI over time to as close to zero as possible. Whether “zero” is attainable, even for individual domains of harm, remains controversial ; however, many now agree that unless the health care industry and hospital systems have “zero” as an aspirational goal, we will never know what is achievable.

Although the PHI was not originally envisioned for comparison between hospitals, it is now being used as a measure of aggregate harm reduction among collaborating hospitals. Groups of hospitals working together using improvement science methods, sharing best practices, and adopting common harm domain definitions, can create regional, state, or even National Harm Indices that can be used as the global measure of harm reduction across large and disparate populations. Variations of the PHI have now been adopted by the Ohio Children’s Hospitals Solutions for Patient Safety, a statewide collaborative of all 8 Ohio children’s hospitals, with planned national expansion with an additional 50 children’s hospitals.