Voluntary Anonymous Reporting of Medical Errors for Neonatal Intensive Care

METHODS

Error Reporting in Network Collaboratives

In October 2000, the Internet site www.nicq.org was enhanced to support voluntary anonymous reporting of medical errors and near-miss errors through hypertext markup language forms. Access to the site was password protected and was limited to prespecified IP addresses at the participating institutions. Overall, 739 health care providers (physicians, nurses, respiratory therapists, pharmacists, and others) from a total of 54 neonatal intensive care units (NICUs) participating in NICQ2000 and NICQ2002 were provided access to the error-reporting site. At the collaborative meetings and via e-mail discussion lists (“listserv”), Vermont Oxford Network faculty encouraged providers to use this web site to report anonymously errors that resulted in harm to the patient as well as near misses. Authorized providers could report events that occur in the NICU, step-down unit, well-infant newborn nursery, delivery room, newborn resuscitation room, mother’s hospital room, other hospital inpatient unit, and operating room, as well as events that occur during interhospital transport of newborn infants. In addition to submitting new error reports, they could search, browse, and view all previously posted reports.

Anonymity of error reporters was ensured by not collecting any information about the individual reporting the error, patient, hospital, date, time, or people involved in the event. In addition, the standard Internet Information Server (Microsoft Corp, Redmond, WA) web log was inactivated, ensuring that no record of the IP address of the submitting computer was stored at the central reporting site.

Phase 1 of Error Reporting

The initial online form, used for error reporting from October 4, 2000, to March 7, 2002 (17 months), contained 4 text boxes labeled Brief Title, Description, Key Words, and References. Each box allowed reporters to type in free text. In addition, reporters were requested to choose keywords relevant to each reported error from a drop-down list. The level of detail contained in the report was left to the reporting individual.

Phase 2 of Error Reporting

A structured online reporting form was introduced on March 8, 2002. The top section of the scroll-down form is shown in Fig 1. This new interface provided several predefined fields for 1) degree of patient harm that resulted from the error, 2) location of the error or near miss, 3) time elapsed between occurrence of the error and submission of the report, 4) categories in which the error would fit, 5) factors that contributed to the event, 6) factors that mitigated the effects of the event, 7) name of the medication if a medication error occurred, and 8) changes implemented to prevent a recurrence of the event. In addition, a text box was provided for feedback and suggestions on how to improve the error reporting system. Responses to fields 1 to 5 in the list above were to be chosen from either a drop-down list or a check box. The available response choices in each of these 5 fields are listed in Appendix 1. Responses to the remaining fields were entered as free text. Reporters had to choose 1 of the following categories to describe the degree of patient harm resulting from the event: event did not have the potential to cause harm, event had the potential to cause harm but did not reach the patient, event reached the patient but did not cause harm, minor harm (increased monitoring, treated with intervention, etc), serious harm (life-threatening, impaired outcome, etc), resulted in death, unknown. This classification was based on the system used for medication errors by the National Coordinating Council for Medication Error Reporting and Prevention.¹⁹

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Top section of the form used for anonymous Internet-based reporting in phase 2.

RESULTS

Of the 1256 reports received during phases 1 and 2 of the study, 26 reports were excluded (22 duplicate reports and 4 reports not related to errors), leaving 1230 reports available for analysis: 522 from phase 1 (17 months) and 708 from phase 2 (10 months). In phase 1, reports varied widely in the level of detail provided, from brief 1-line statements to paragraphs rich in detail. They did not consistently include the occurrence or lack of patient harm with a reported event.

The distribution of reported events from phases 1 and 2 across the 4 major categories and subcategories of the modified Leape’s classification²⁰ is shown in Table 1. Illustrative examples of the broad range of reported events from phases 1 and 2 are provided in Table 2. There was a high level of concordance between the 2 reviewers in this classification (κ = .82). Approximately half (581 [47%]) of all reported events from phases 1 and 2 were related to medications, nutritional agents (breast milk, formula, and parenteral nutrition), or blood products. The distribution of these 581 reports across the stages of the medication system is shown in Table 3. Thirty-one percent of errors occurred at the administration stage, 25% at the dispensing stage, 16% at the ordering stage, and 12% at the transcribing stage. Only 1.4% were errors of drug monitoring. There was a high level of concordance between the 2 reviewers (κ = .98) for their classifications of the stages. When reports from phases 1 and 2 involving patient misidentification from the “diagnosis,” “treatment,” “prevention,” and “other” categories in Table 1 were combined, they composed 11.1% (n = 136) of all reports.

Of all of the reports in phases 1 and 2, there were 24 in which a parent or family member was involved in the event. In 7 reports, a parent contributed to the occurrence of the error (eg, administering a medication that was not ordered, incorrectly mixing nutrients for feeding). In 8 reports, the family member was the affected individual (eg, falling as a result of a hazardous condition in the environment, not being informed of a procedure or patient transfer). In 9 reports, a parent was described as either discovering an error or participating in investigating the cause.

Of the 707 reports in phase 2 in which the location of occurrence of the event was reported, the location was the NICU in 675 (95.5%), the intermediate care or step-down unit in 20 (2.8%), the delivery room in 3 (0.42%), during transport from another hospital in 3 (0.42%), other hospital inpatient unit in 2 (0.28%), the mother’s hospital room in 1 (0.14%), and the operating room in 1 (0.14%). Two events were reported to have occurred in other locations, which were not named by reporting individuals. In 1 report, the location field was left blank. No events were reported from the well-infant newborn nursery, the newborn resuscitation room, or the radiology department. No events were reported to have occurred during transport to another hospital or during transport within the hospital.

Among the 686 phase 2 reports that contained the interval between event and reporting, the time elapsed was >28 days in 346 (50.4%), 8 to 28 days in 144 (21%), 4 to 7 days in 67 (9.8%), 1 to 3 days in 44 (6.4%), 4 to 24 hours in 54 (7.9%), and <4 hours in 31 (4.5%).

Among the 584 (82.5%) phase 2 reports in which at least 1 contributory factor was reported, 1 contributory factor was selected in 216 (37%) reports, 2 contributory factors in 170 (29.1%), 3 factors in 98 (16.8%), and 4 factors in 47 (8%). In 52 (8.9%)reports, 5 to 8 factors were selected per report. In 1 report, 11 factors were selected. Table 4 shows the types of contributory factors that were reported in phase 2. The most frequent contributory factors were failure to follow policy or protocol (47%), inattention (27%), communications problem (22%), error in charting or documentation (13%), distraction (12%), inexperience (10%), labeling error (10%), and poor teamwork (9%).

The degree of harm resulting from the 673 events in which this outcome was reported during phase 2 is shown in Fig 2. Actual harm to the patient occurred in 181 (27%) of the reported events, with minor harm (ie, requirement for increased monitoring, specific treatment, other interventions) resulting in 167 (25%) cases, serious harm in 13 (1.9%) cases, and death in 1 (0.2%) case.

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

Categories of harm for events reported in phase 2. Outcome of 673 events in which degree of harm was reported. This outcome was not reported in 35 events. Minor harm, which consisted of increased monitoring, treated with intervention, etc, occurred in 167 (25%) events. Serious harm, in which there was a threat to life or an impaired outcome, etc, was reported in 13 (1.9%) events. Death was reported in 1 (0.15%) event. Categories are based on the system used for medication errors by the National Coordinating Council for Medication Error Reporting and Prevention.¹⁹

Periodic structured feedback of collected errors, both through the NICQ 2002 collaborative’s e-mail listserv and at biannual meetings, has increased awareness of errors and stimulated numerous patient safety improvement projects. The 739 health professionals (including physicians, nurses, respiratory therapists, pharmacists, and administrators) who had access to this web site represented a broad range of disciplines. Discussion of reported errors has provided a strong impetus for multidisciplinary learning. Multidisciplinary teams for hospitals in the improvement collaborative have prepared 141 poster presentations of case studies on patient safety in their units and shared these at meetings of the collaborative. Teams have presented the results of 43 root-cause analyses, including insulin overdoses, accidental dislodgment of intravenous catheters, administration of wrong medications, falls, breast milk errors, and tracheal perforations. Teams have also used failure mode and effects analysis (FMEA) and hazard analysis and critical control point methods, notably for improving systems for processing and feeding breast milk to infants.

DISCUSSION

Voluntary reporting of medical errors and adverse events is a key component of patient safety efforts, but health care providers may not report such events for fear of being stigmatized, punished by superiors, or exposed to medical-legal liability.^10,23–25 The success of the Vermont Oxford Network Internet reporting system, NICQ.org, demonstrates that it is possible to overcome such barriers. Health care providers will voluntarily report significant medical errors and adverse events to an external organization using a safe, convenient, specialty-based system—especially when there is trust in the organization and a strong foundation of multidisciplinary collaborative quality improvement such as that within our network.

This study provides data on the broad range of medical errors and adverse events in neonatal intensive care. Our experience suggests that medical errors are a major problem in neonatal intensive care, as they are in other areas of health care. Of the reported errors, 27% resulted in harm to the patient. Errors occurred in a wide variety of domains, such as medications, diagnostic testing, medical or surgical procedures, respiratory therapy, the environment of care, patient identification, nursing care, and physician–patient communication (Tables 1 and 2). A number of reported errors pertained to the safety goals promulgated recently by the JCAHO,²⁶ specifically the use of high-alert medications, patient identification, and communication among caregivers. Approximately half of all reports involved medications, nutritional products, or blood products, exceeding rates published previously from pediatric²⁷ and adult^12,28–30 ICUs in which 12% to 44% of reports involved medication errors. Previous studies that were restricted to medication errors^22,31–33 documented a high frequency of such errors in NICUs, and a prospective pediatric study using active error detection methods found that the most serious errors occurred significantly more often in the NICU than on other wards.²²

The Agency for Healthcare Research and Quality and the American Academy of Pediatrics encourage parents to be actively involved in error prevention in both ambulatory care and inpatient settings.³⁴ Our findings demonstrate that family members of a hospitalized neonate not only can help discover errors but also can contribute to the cause of the error or themselves be victims of the error. Additional research is required to understand how best to involve families in efforts to enhance the safety of medical care for their hospitalized infants.

In addition to identifying errors and adverse events, in phase 2, we attempted to identify the factors that contributed to their occurrence. Forty-seven percent of reports were associated with a failure to follow a policy or protocol, 27% with inattention, 22% with a communication problem, and 12% with distraction. In contrast, problems in communication, orientation/training, and patient assessment were the 3 most frequent root causes of sentinel events reported to JCAHO.³⁵ We recognize that voluntary reports can provide only a glimpse into the complex cause of error, but our findings suggest that merely issuing new policies or protocols will not improve patient safety in neonatal intensive care. Instead, greater success may result from system-wide improvements that decrease inattention, miscommunication, distraction, and other contributory human factors. These principles are implicit in contemporary “systems” approaches to understanding error and improving safety,^36–39 including root-cause analysis,⁴⁰ FMEA,⁴¹ hazard analysis and critical control point,⁴² and crew resource management.^43,44 Surprising, nurse understaffing was infrequently (2%) reported as a contributory factor, although several studies have discovered an association between high patient–nurse ratios and adverse patient outcomes.^45–51

Our study was limited in several ways. First, because reporting was voluntary, the frequency of events reported does not represent the true incidence of errors in participating NICU. All voluntary systems are likely to suffer from incomplete ascertainment and selective reporting. Prospective active surveillance systems, of which the National Nosocomial Infections Surveillance System⁶ is a good model, are required to determine the true incidence of adverse events. Although such systems are critical for understanding the epidemiology of adverse events, they are labor intensive and not well suited for capturing errors that are not documented in medical records or cannot be inferred from laboratory or pharmacy data. Second, we suspect that the intensity and comprehensiveness of local investigation of errors varied widely, and the reliability with which contributing factors were identified probably varied as well. In future versions of our reporting system, we will ascertain whether contributing factors were identified through formal root-cause analysis. Finally, we were unable to identify the number of reports submitted by individual NICUs or by members by professional discipline because reporting was fully anonymous. This is a disadvantage of anonymous reporting, as is the inability to contact reporting individuals for details of a reported event.¹

Several benefits of multi-institutional specialty-based voluntary reporting systems such as the Vermont Oxford Network system are evident. They can identify errors that are rare in individual institutions but occur in multiple institutions. For example, on www.nicq.org, there were several reports of intravenous infusion of solutions intended for enteral use—a systems problem that can be prevented by designing enteral tubing and syringes that cannot be connected to intravenous tubing (a so-called “physical constraint”). Multi-institutional specialty-based reporting systems also can identify patterns of errors that are unique to that specialty. For example, there were multiple reports of infants’ being fed breast milk from the wrong mother and of infants actually or nearly falling from an incubator, weighing scale, or swing while being cared for in the NICU.

According to Leape,² successful reporting systems are nonpunitive, confidential, and independent; provide expert analysis and timely feedback; and are systems oriented and responsive. The Vermont Oxford Network reporting system already adheres at least partially to all of these criteria. An expert faculty analyzes reported events and provides feedback and recommendations to all participating NICUs by e-mail and at the collaborative’s biannual meetings. The next step in the development of nicq.org is to transform it from a reporting system to a real-time learning system. Individuals who report errors will receive immediate electronic feedback on the web site and direct links to case studies of similar errors, prevention tips, and patient safety resources such as root-cause analysis and FMEA tools and human factors checklists. The long-term vision is that shared learning among the universe of NICUs will enhance patient safety.

CONCLUSIONS

We have successfully developed and implemented a simple, anonymous, voluntary, Internet-based reporting system for medical errors in neonatal intensive care. We speculate that similar voluntary reporting systems have the potential to enhance patient safety in a wide range of medical and surgical specialties.

APPENDIX 1: LISTS OF RESPONSE CHOICES FOR CATEGORY OF HARM, LOCATION OF EVENT, TIME SINCE EVENT, AND CATEGORIES OF ERROR IN THE WWW.NICQ.ORG INTERNET REPORTING FORM USED IN PHASE 2

Actual or potential harm: did the event cause harm (please choose one):

1. No, event did not have the potential to cause harm

2. No, event had potential to cause harm but did not reach patient

3. No, event reached patient but did not cause harm

4. Yes, minor harm (increased monitoring, treated with intervention, etc)

5. Yes, serious harm (life-threatening, impaired outcome, etc)

6. Yes, resulted in death

7. Unknown

Where was the patient when the error or near miss occurred? Location (please choose one):

1. Neonatal intensive care unit

2. Intermediate care or step-down unit

3. Well-infant newborn nursery

4. Delivery room

5. Newborn resuscitation room

6. Mother’s hospital room

7. Other hospital inpatient unit

8. During transport from another hospital

9. During transport to another hospital

10. During transport within your hospital

11. Operating room

12. Radiology department

13. Other

How long ago did the event occur? Time since event (please choose one):

1. <4 hours

2. 4 to <24 hours

3. 1 to 3 days

4. 4 to 7 days

5. 8 to 28 days

6. >28 days

7. Unknown

Categories of errors (check all that apply)

• Medication or drug

• Breast milk

• Enteral feeding other than breast milk

• Parenteral nutrition

• Central line or vascular access

• Infusion or infiltrate

• Fluid or electrolyte

• Anesthesia, analgesia, or sedation

• Respiratory care or ventilator

• Glucose or insulin

• Monitoring or alarms

• Radiology or diagnostic imaging

• Surgery

• Transportation in or between hospitals

• Transfusion

• Laboratory testing

• Family or visitors

• Security

• Patient misidentification

• Informed consent

• Resuscitation

• Medical devices or equipment

• Enter any other categories of errors that apply

Factors that contributed to the event (check all that apply)

Environment

• Inadequate lighting

• Lack of space or room

• Noise

• Unfamiliar environment

Equipment

• Equipment failure

• Inadequate equipment maintenance

• Necessary equipment unavailable

• Poor equipment design

• Unfamiliar equipment

Human factors

• Confrontational or intimidating behavior

• Distraction

• Fatigue

• Inadequate training

• Inattention

• Inexperience

• Stress

Practices

• Calculation error

• Communications problem

• Error in charting or documentation

• Error in computer entry

• Failure to follow policy or protocol

• Inadequate protocol

• Inadequate security

• Labeling error

• Lack of supervision

• Nursing handoff or shift change

• Physician handoff or shift change

• Poor teamwork

• Inability to contact needed staff

• Wrong protocol used

Staffing

• Consultant or subspecialist unavailable

• High census in unit

• High patient acuity in unit

• Low levels of clerical or support staff

• Low levels of other professional staff

• Low nursing staff levels

• Low physician staff levels

APPENDIX 2: NICQ 2000 AND 2002 PARTICIPANTS