- Email: [email protected]
Synchronization of Timepieces to the Atomic Clock in an Urban Emergency Medical Services System
EMS/BRIEF REPORT
Synchronization of Timepieces to the Atomic Clock in an Urban Emergency Medical Services System From the Department of Emergency Medicine, Virginia Commonwealth University/Medical College of Virginia,* Mercy Ambulance of Richmond,‡ Richmond Ambulance Authority,§ Richmond Division of Emergency Communications, Richmond, VA.II Received for publication January 23, 1997. Revisions received October 8, 1997, and November 19, 1997. Accepted for publication December 2, 1997.
Joseph P Ornato, MD* Mark L Doctor‡ Lori F Harbour, REMTP‡ Mary Ann Peberdy, MD* Jerry Overton, MPA§ Edward M Racht, MD* William G Zauhar‡ Alan P Smith‡ Kent A RyanII
Presented in poster format at the American Heart Association Scientific Sessions, 1994. Copyright © 1998 by the American College of Emergency Physicians.
Study objective: Erroneous time documentation of emergency treatment caused by the variation in the accuracy of timepieces has profound medical, medicolegal, and research consequences. The purpose of this study was to confirm the variation of critical timepiece settings in an urban emergency care system noted in previous studies and to implement and monitor the results of a prospective program to improve time synchronization. Methods: Timepieces (n=393) used by firefighters, paramedics, and emergency physicians and nurses were randomly sampled immediately before and at two time intervals (1 and 4 months) after attempted synchronization to the US atomic clock standard. The setting on each timepiece was compared with the atomic clock. From the data, a mathematical simulation estimated the number of time-related documentation errors that would occur in 2,500 simulated cardiac arrest cases using timepieces with accuracy similar to those found in the EMS system before and after attempted synchronization. Results: Before attempted synchronization, the timepieces had a mean error of 2.0 (95% confidence interval 1.8 to 2.3) minutes. One month after attempted synchronization, the mean error decreased significantly to .9 (.8 to 1.1) minute. However, it increased to 1.7 (1.5 to 1.9) minutes within 4 months. Mathematical simulation before attempted synchronization predicted that 93% of cardiac arrest cases would contain a documentation error of 2 minutes or more and that 41% of cases would contain a documentation error of 5 minutes or more. Attempted synchronization cut the 2minute documentation error rate in half and reduced the 5-minute documentation error rate by three fourths. However, the error rates were predicted to return to baseline 4 months after attempted synchronization. Conclusion: Emergency medical timepieces are often inaccurate, making it difficult to reconstruct events for medical, medicolegal, or research purposes. Community synchronization of timepieces to the atomic clock can reduce the problem significantly, but
APRIL 1998
31:4
ANNALS OF EMERGENCY MEDICINE
4 8 3
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
the effects of a one-time attempted synchronization event are short-lived. [Ornato JP, Doctor ML, Harbour LF, Peberdy MA, Overton J, Racht EM, Zauhar WG, Smith AP, Ryan KA: Synchronization of timepieces to the atomic clock in an urban emergency medical services system. Ann Emerg Med April 1998;31:483-487.]
INTRODUCTION
Accurate recording of the time that a critical patient care event occurs is vital for clinical, medicolegal, quality improvement, and research purposes. This is especially true for practitioners who frequently treat patients needing multiple, quickly accomplished interventions. If only one timepiece is used, the sequence of events and the intervals between events should be accurate. Use of multiple, unsynchronized timepieces to record clinical events for the same patient increases the likelihood that significant time documentation errors will occur. Several authors have noted that the problem exists,1-5 but only one EMS system has attempted to quantify its magnitude.5 The purpose of this study was to confirm the variation of critical timepieces in an urban EMS system and to prospectively implement and monitor the results of a program to improve time synchronization. M AT E R I A L S A N D M E T H O D S Study site EMS system
The study was conducted in the City of Richmond, Virginia (population 260,000), which has a modified public utility model EMS system. Fire department EMTs equipped with automated external defibrillators (AEDs) and other first-aid equipment provide first-response stabilization and treatment. Richmond’s 26 ambulances provide advanced life support. Data sampling
A research paramedic (LFH) sampled all of the time data points reported in the study. The computer-aided dispatch system in both the 911 center and the emergency medical dispatch center are linked together electronically. Both systems are synchronized daily by automatic telephone modem linkage to the US Department of Commerce “gold standard” atomic clock in Boulder, Colorado. The study was conducted in three phases. During each phase, the research paramedic sampled timepieces used by Richmond firefighters, paramedics, and the physicians and nurses staffing each of the city’s seven EDs. The research
4 8 4
paramedic compared the time readings on each emergency care timepiece with an analog quartz crystal-controlled watch that was set and checked several times daily to ensure that it was synchronized with the atomic clock standard. The analog watch was set at the start of each day by calling a local telephone number that indicated the correct time every 10 seconds. The analog watch did not at any time deviate more than 1 second from the Boulder atomic clock reading in any given day during each phase of data collection. Timepieces that were sampled included watches worn or carried by firefighters, paramedics, emergency nurses, emergency physicians, and timers on AEDs and manual defibrillators on fire trucks, ambulances, and in ED “code rooms.” Phase 1 was a baseline control period. After phase 1, a one-time program was implemented to synchronize as many of the timepieces in the prehospital and ED environment as possible. One month after this single attempted synchronization event, a phase 2 random sampling of the same kind of timepieces was performed by the research paramedic using the same method. Four months after the attempted synchronization, a final phase 3 random sampling was conducted. The attempted synchronization event consisted of a onetime simultaneous resetting of all timepieces to the atomic clock standard. The research paramedic personally visited each of the area EDs, fire stations, and paramedic headquarters at least once (and in many cases, multiple times) in the week before attempted synchronization. All personnel (physicians, nurses, firefighters, and paramedics) were instructed on how to synchronize their timepieces by calling the local telephone number that announces the atomic clock time every 10 seconds. They were asked to call the number on a specific day at 8 AM. Reminders were given for several days before the attempted synchronization day. Finally, the research paramedic met with each area hospital building maintenance team that had responsibility for resetting ED “code room” clocks. The research paramedic explained the purpose of the program and requested that the maintenance teams reset the “code room” clocks to the atomic clock standard at 8 AM on the assigned attempted synchronization day. The research paramedic also met with and trained fire and paramedic maintenance personnel who were responsible for resetting the AEDs and the manual paramedic defibrillators. Data analysis
The error of each timepiece sampled during each of the three phases was measured and entered into a database (Access version 2.0, Microsoft, Redmond, WA). Timepieces that were slower than the atomic clock standard were recorded as a negative value; timepieces that were faster than
ANNALS OF EMERGENCY MEDICINE
31:4
APRIL 1998
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
the atomic clock standard were recorded as a positive value. Tables were created that contained the actual time error readings for each of the timepiece categories. Separate tables were created that consisted of the absolute value of the timepiece errors (indicating the number of minutes that each timepiece erred from the atomic clock standard without regard to whether the device was actually fast or slow). The absolute value tables were used to calculate the mean (95% confidence interval [CI]) timepiece error in each phase of the study for all timepieces as a group, and for each of six timepiece categories (firefighter watches, firefighter defibrillators, paramedic watches, paramedic defibrillators, emergency physician and nurse watches, and ED wall clocks). A mathematical simulation was created to model what would happen if the timepieces from each phase of the study were used to document the sequence of events on 2,500 cardiac arrests (which would represent approximately 5 years of clinical resuscitation activity in our EMS system). It was assumed that the error on each timepiece would remain constant during the mathematical simulation period. Each simulation “run” was created by having the computer randomly select an error reading from each of five EMS timepiece categories using data from the actual (rather than the absolute value) timepiece tables. ED wall clocks and watches worn by physicians and nurses were all considered as a single group for purposes of the simulation, because the relative use of each of these timepieces for documentation during a cardiac arrest was unknown. The computer determined the maximum error that would occur on each run by subtracting the error of the timepiece with the slowest reading from the error of the timepiece with the fastest reading relative to the actual time. After the 2,500 simulated runs were generated and stored in a table, the computer calculated the percentage of simulated runs
in which there would be a 2- or a 5-minute recording error present based on timepiece variability. The simulation was run on data collected from phases 1, 2, and 3 separately. Statistical analysis was performed using Statistica version 5.0 (Statsoft, Inc, Tulsa, OK). Because taking the absolute value of time measurements results in numbers that are not normally distributed, the Kruskal-Wallis test was used for analysis, followed by the Nemenyi test for nonparametric multiple comparisons.6 R E S U LT S
A total of 1,192 timepiece measurements were made in all three phases combined. Thirty-nine of these measurements were excluded from analysis because they were more than 15 minutes faster or slower than the atomic clock. These values were excluded because it was reasoned that they were sufficiently deviant from the correct time that health care providers would likely realize that something might be wrong with the timepiece and not use it. This excluded only 3% of the measurements, resulting in normally distributed curves for each of the timepieces in each of the phases. Altering the exclusion limit to 20 or 30 minutes did not significantly alter the results. The mean timepiece errors grouped by timepiece category and study phase are shown in Table 1. Prior to attempted synchronization, the timepieces had a mean error of 2.0 (95% CI 1.8 to 2.3) minutes. One month after the attempted synchronization the mean error decreased significantly to .9 (.8 to 1.1) minute. By 4 months, the mean error had reverted to 1.7 (1.5 to 1.9) minutes. The greatest improvement in timepiece synchronization at 1 month was noted in firefighter AEDs, paramedic watches, paramedic defibrillators, and ED wall clocks. The
Table 1.
Actual data—absolute value of the error of timepieces in minutes before (phase 1), 1 month (phase 2), and 4 months (phase 3) after a one-time community-attempted synchronization event. Timepiece
Phase 1 (No., 95% CI)
Firefighter wristwatches Firefighter AEDs Paramedic wristwatches Paramedic defibrillators Emergency nurse and physician watches ED wall clocks Totals
1.82 (69, 1.25–2.38)* .74 (28, .19–1.29)* 1.20 (42, .79—1.62)* 3.12 (53, 2.44–3.81)* 1.83 (120, 1.51–2.15)* 2.73 (81, 2.15—3.30)* 2.04 (393, 1.82–2.26)*
Phase 2 (No., 95% CI) 1.08 (64, .81–1.35)*† .01 (30, .01–.02)*† .49 (40, .31–.67)*† .84 (51, .26–1.41)*† 1.40 (119, 1.11–1.70)*† .76 (76, .52–1.00)*† .94 (380, .80–1.08)*†
Phase 3 (No., 95% CI) 1.49 (67, 1.03—1.95)† .59 (30, .36–.81)† .78 (33, .44–1.13)† 2.26 (50, 1.60–2.93)† 1.90 (126, 1.51–2.29)† 2.11 (74, 1.45–2.77)† 1.72 (380, 1.49–1.94)†
Results presented as mean number of minutes timepieces deviated from the actual atomic clock time. *P<.05. †P<.05.
APRIL 1998
31:4
ANNALS OF EMERGENCY MEDICINE
4 8 5
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
least improvement was noted in watches worn by firefighters, emergency nurses, and emergency physicians. The error of all timepieces was not statistically significantly different from baseline in any timepiece category 4 months after the one-time attempted synchronization event. Simulation of cardiac arrest cases based on random sampling of the timepieces (Table 2) predicted that, before attempted synchronization, 93% of cases would contain a documentation error of 2 minutes or more and 41% of cases would contain a documentation error of 5 minutes or more. At 1 month, the synchronization event cut the 2-minute documentation error rates almost in half, and reduced the 5-minute documentation error rate by more than three fourths. The error rates nearly returned to baseline by 4 months after the attempted synchronization event. DISCUSSION
Our study confirms that serious documentation errors can result from lack of time synchronization in an urban EMS system. However, unlike previous investigations, we quantified the effect and short-lived benefits of a communityattempted synchronization event. Our results confirm the need for a continuous program of time synchronization in an urban EMS system. Investigators studying similar systems have found that time synchronization is a common problem and the subject has been the focus of editorial commentary.1-5 Cordell et al4 found an average 1-minute 45-second–absolute difference (range of 12 minutes 34 seconds slow to 7 minutes 7 seconds fast) between timepieces sampled on a single day in Indianapolis EDs, ambulances, and fire vehicles and the atomic clock standard. Becker et al2 documented time discrepancies in 39% of prehospital cardiac arrest cases in Chicago. Others have noted significant differences between Table 2.
Simulation data—percent of cases with ≥2- and ≥5-minute documentation errors that would be present on documents recording events on simulated cardiac arrest cases using timepieces with accuracy determined during phases 1, 2, and 3.
No. Cases with ≥2-minute documentation error (%) Cases with ≥5-minute documentation error (%)
4 8 6
Phase 1
Phase 2
Phase 3
2,500 93
2,500 52
2,500 83
41
10
23
paramedic “trip report” times and actual time determined from real-time audio recordings during the runs.1,3 Our investigation included a much larger number and variety of timepieces than those sampled by other investigators. We also explored the implications of inaccurate timepieces by mathematically simulating the cardiac arrest documentation errors that would occur if prehospital cardiac arrest records were generated by personnel using the timepieces to document emergency medical care during resuscitation. Our findings raise serious questions about the medical and medicolegal accuracy of time-based documentation during life-or-death emergency medical treatment. The use of erroneous timepieces can have devastating effects on the documentation of resuscitation emergencies such as those caused by cardiac arrest or major trauma. In these conditions, documentation usually involves multiple timepieces and a large number of events occur per unit of time. This makes it difficult or impossible to reconstruct events accurately for medical, quality improvement, medicolegal, or research purposes. EMS research that is based on the time sequence of events should take into account the special challenges faced by the problem of time synchronization. It seems appropriate for researchers to describe the way in which time is measured in the EMS system, the techniques that are used to standardize timepieces, and the amount of variability that was present among timepieces during the study period. These details should be added to the uniform elements of data collection already recommended for events such as cardiac arrest.5 Ideally, all clocks should be set universally by a central server. For clocks that are controlled by a computer, there are numerous inexpensive programs available on publicly accessible software networks that will command the computer to automatically dial and download the correct time into its timepiece from the atomic clock on a daily basis. Limitations of this study include the fact that only timepieces used to create the records, rather than actual records, were checked for accuracy. In addition, there was no way to “guarantee” that all timepieces were perfectly synchronized immediately after the attempted synchronization event. Another study limitation is that it was conducted in an EMS system that is noted for its excellence in clinical care and continuous quality improvement program. As such, it is likely that the results obtained may actually underestimate the extent and severity of the problem of time synchronization in other, less advanced systems. In addition, the investigation did not sample timepieces in areas of the hospital other than the ED. Thus it would be improper to
ANNALS OF EMERGENCY MEDICINE
31:4
APRIL 1998
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
extrapolate the findings of this study to other clinical areas. Another potential limitation is that the study focused on the “worst-case” scenario, in which health care workers do not realize that a time documentation error has occurred and do not attempt to adjust for it or account for it in their narrative report. In actuality, some time documentation errors are discovered by health care workers and corrected or discussed in the medical record. Results of this study indicate that timepieces used by EMS providers are often inaccurate, making it difficult to reconstruct events for medical, medicolegal, or research purposes. A community synchronization of timepieces to the atomic clock can reduce the problem significantly, but the effects of a one-time attempted synchronization event are short-lived.
Reprint no. 47/1/88737 Address for reprints: Joseph P Ornato, MD Department of Emergency Medicine Medical College of Virginia 401 North 12th Street, Room G248 Box 980525 Richmond, VA 23298-0525 804-828-7184 Fax 804-828-8597
REFERENCES 1. Mosesso VN Jr: The most neglected tool in EMS: The clock. Ann Emerg Med 1993;22:1311-1312. 2. Becker LB, Ostrander MP, Barrett J, et al: Outcome of CPR in a large metropolitan area— Where are the survivors? Ann Emerg Med 1991;20:356-361. 3. Spaite DW, Valenzuela TD, Meislin HW, et al: Prospective validation of a new model for evaluating emergency medical services systems by in-field observation of specific time intervals in prehospital care. Ann Emerg Med 1993;22:638-645. 4. Cordell WH, Olinger ML, Kozak PA, et al: Does anybody really know what time it is? Does anybody really care? Ann Emerg Med 1994;23:1032-1036. 5. Cummins RO, Chamberlain DA, Abramson NS, et al: Recommended guidelines for uniform reporting of data from out-of-hospital cardiac arrest: The Utstein style. A statement for health professions from a task force of the American Heart Association, the European Resuscitation Council, the Heart and Stroke Foundation of Canada, and the Australian Resuscitation Council. Circulation 1991;84:960-975. 6. Zar JH: Biostatistical Analysis, ed 2. Englewood Cliffs, NJ: Prentice-Hall, 1994:199-201.
APRIL 1998
31:4
ANNALS OF EMERGENCY MEDICINE
4 8 7
Synchronization of Timepieces to the Atomic Clock in an Urban Emergency Medical Services System From the Department of Emergency Medicine, Virginia Commonwealth University/Medical College of Virginia,* Mercy Ambulance of Richmond,‡ Richmond Ambulance Authority,§ Richmond Division of Emergency Communications, Richmond, VA.II Received for publication January 23, 1997. Revisions received October 8, 1997, and November 19, 1997. Accepted for publication December 2, 1997.
Joseph P Ornato, MD* Mark L Doctor‡ Lori F Harbour, REMTP‡ Mary Ann Peberdy, MD* Jerry Overton, MPA§ Edward M Racht, MD* William G Zauhar‡ Alan P Smith‡ Kent A RyanII
Presented in poster format at the American Heart Association Scientific Sessions, 1994. Copyright © 1998 by the American College of Emergency Physicians.
Study objective: Erroneous time documentation of emergency treatment caused by the variation in the accuracy of timepieces has profound medical, medicolegal, and research consequences. The purpose of this study was to confirm the variation of critical timepiece settings in an urban emergency care system noted in previous studies and to implement and monitor the results of a prospective program to improve time synchronization. Methods: Timepieces (n=393) used by firefighters, paramedics, and emergency physicians and nurses were randomly sampled immediately before and at two time intervals (1 and 4 months) after attempted synchronization to the US atomic clock standard. The setting on each timepiece was compared with the atomic clock. From the data, a mathematical simulation estimated the number of time-related documentation errors that would occur in 2,500 simulated cardiac arrest cases using timepieces with accuracy similar to those found in the EMS system before and after attempted synchronization. Results: Before attempted synchronization, the timepieces had a mean error of 2.0 (95% confidence interval 1.8 to 2.3) minutes. One month after attempted synchronization, the mean error decreased significantly to .9 (.8 to 1.1) minute. However, it increased to 1.7 (1.5 to 1.9) minutes within 4 months. Mathematical simulation before attempted synchronization predicted that 93% of cardiac arrest cases would contain a documentation error of 2 minutes or more and that 41% of cases would contain a documentation error of 5 minutes or more. Attempted synchronization cut the 2minute documentation error rate in half and reduced the 5-minute documentation error rate by three fourths. However, the error rates were predicted to return to baseline 4 months after attempted synchronization. Conclusion: Emergency medical timepieces are often inaccurate, making it difficult to reconstruct events for medical, medicolegal, or research purposes. Community synchronization of timepieces to the atomic clock can reduce the problem significantly, but
APRIL 1998
31:4
ANNALS OF EMERGENCY MEDICINE
4 8 3
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
the effects of a one-time attempted synchronization event are short-lived. [Ornato JP, Doctor ML, Harbour LF, Peberdy MA, Overton J, Racht EM, Zauhar WG, Smith AP, Ryan KA: Synchronization of timepieces to the atomic clock in an urban emergency medical services system. Ann Emerg Med April 1998;31:483-487.]
INTRODUCTION
Accurate recording of the time that a critical patient care event occurs is vital for clinical, medicolegal, quality improvement, and research purposes. This is especially true for practitioners who frequently treat patients needing multiple, quickly accomplished interventions. If only one timepiece is used, the sequence of events and the intervals between events should be accurate. Use of multiple, unsynchronized timepieces to record clinical events for the same patient increases the likelihood that significant time documentation errors will occur. Several authors have noted that the problem exists,1-5 but only one EMS system has attempted to quantify its magnitude.5 The purpose of this study was to confirm the variation of critical timepieces in an urban EMS system and to prospectively implement and monitor the results of a program to improve time synchronization. M AT E R I A L S A N D M E T H O D S Study site EMS system
The study was conducted in the City of Richmond, Virginia (population 260,000), which has a modified public utility model EMS system. Fire department EMTs equipped with automated external defibrillators (AEDs) and other first-aid equipment provide first-response stabilization and treatment. Richmond’s 26 ambulances provide advanced life support. Data sampling
A research paramedic (LFH) sampled all of the time data points reported in the study. The computer-aided dispatch system in both the 911 center and the emergency medical dispatch center are linked together electronically. Both systems are synchronized daily by automatic telephone modem linkage to the US Department of Commerce “gold standard” atomic clock in Boulder, Colorado. The study was conducted in three phases. During each phase, the research paramedic sampled timepieces used by Richmond firefighters, paramedics, and the physicians and nurses staffing each of the city’s seven EDs. The research
4 8 4
paramedic compared the time readings on each emergency care timepiece with an analog quartz crystal-controlled watch that was set and checked several times daily to ensure that it was synchronized with the atomic clock standard. The analog watch was set at the start of each day by calling a local telephone number that indicated the correct time every 10 seconds. The analog watch did not at any time deviate more than 1 second from the Boulder atomic clock reading in any given day during each phase of data collection. Timepieces that were sampled included watches worn or carried by firefighters, paramedics, emergency nurses, emergency physicians, and timers on AEDs and manual defibrillators on fire trucks, ambulances, and in ED “code rooms.” Phase 1 was a baseline control period. After phase 1, a one-time program was implemented to synchronize as many of the timepieces in the prehospital and ED environment as possible. One month after this single attempted synchronization event, a phase 2 random sampling of the same kind of timepieces was performed by the research paramedic using the same method. Four months after the attempted synchronization, a final phase 3 random sampling was conducted. The attempted synchronization event consisted of a onetime simultaneous resetting of all timepieces to the atomic clock standard. The research paramedic personally visited each of the area EDs, fire stations, and paramedic headquarters at least once (and in many cases, multiple times) in the week before attempted synchronization. All personnel (physicians, nurses, firefighters, and paramedics) were instructed on how to synchronize their timepieces by calling the local telephone number that announces the atomic clock time every 10 seconds. They were asked to call the number on a specific day at 8 AM. Reminders were given for several days before the attempted synchronization day. Finally, the research paramedic met with each area hospital building maintenance team that had responsibility for resetting ED “code room” clocks. The research paramedic explained the purpose of the program and requested that the maintenance teams reset the “code room” clocks to the atomic clock standard at 8 AM on the assigned attempted synchronization day. The research paramedic also met with and trained fire and paramedic maintenance personnel who were responsible for resetting the AEDs and the manual paramedic defibrillators. Data analysis
The error of each timepiece sampled during each of the three phases was measured and entered into a database (Access version 2.0, Microsoft, Redmond, WA). Timepieces that were slower than the atomic clock standard were recorded as a negative value; timepieces that were faster than
ANNALS OF EMERGENCY MEDICINE
31:4
APRIL 1998
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
the atomic clock standard were recorded as a positive value. Tables were created that contained the actual time error readings for each of the timepiece categories. Separate tables were created that consisted of the absolute value of the timepiece errors (indicating the number of minutes that each timepiece erred from the atomic clock standard without regard to whether the device was actually fast or slow). The absolute value tables were used to calculate the mean (95% confidence interval [CI]) timepiece error in each phase of the study for all timepieces as a group, and for each of six timepiece categories (firefighter watches, firefighter defibrillators, paramedic watches, paramedic defibrillators, emergency physician and nurse watches, and ED wall clocks). A mathematical simulation was created to model what would happen if the timepieces from each phase of the study were used to document the sequence of events on 2,500 cardiac arrests (which would represent approximately 5 years of clinical resuscitation activity in our EMS system). It was assumed that the error on each timepiece would remain constant during the mathematical simulation period. Each simulation “run” was created by having the computer randomly select an error reading from each of five EMS timepiece categories using data from the actual (rather than the absolute value) timepiece tables. ED wall clocks and watches worn by physicians and nurses were all considered as a single group for purposes of the simulation, because the relative use of each of these timepieces for documentation during a cardiac arrest was unknown. The computer determined the maximum error that would occur on each run by subtracting the error of the timepiece with the slowest reading from the error of the timepiece with the fastest reading relative to the actual time. After the 2,500 simulated runs were generated and stored in a table, the computer calculated the percentage of simulated runs
in which there would be a 2- or a 5-minute recording error present based on timepiece variability. The simulation was run on data collected from phases 1, 2, and 3 separately. Statistical analysis was performed using Statistica version 5.0 (Statsoft, Inc, Tulsa, OK). Because taking the absolute value of time measurements results in numbers that are not normally distributed, the Kruskal-Wallis test was used for analysis, followed by the Nemenyi test for nonparametric multiple comparisons.6 R E S U LT S
A total of 1,192 timepiece measurements were made in all three phases combined. Thirty-nine of these measurements were excluded from analysis because they were more than 15 minutes faster or slower than the atomic clock. These values were excluded because it was reasoned that they were sufficiently deviant from the correct time that health care providers would likely realize that something might be wrong with the timepiece and not use it. This excluded only 3% of the measurements, resulting in normally distributed curves for each of the timepieces in each of the phases. Altering the exclusion limit to 20 or 30 minutes did not significantly alter the results. The mean timepiece errors grouped by timepiece category and study phase are shown in Table 1. Prior to attempted synchronization, the timepieces had a mean error of 2.0 (95% CI 1.8 to 2.3) minutes. One month after the attempted synchronization the mean error decreased significantly to .9 (.8 to 1.1) minute. By 4 months, the mean error had reverted to 1.7 (1.5 to 1.9) minutes. The greatest improvement in timepiece synchronization at 1 month was noted in firefighter AEDs, paramedic watches, paramedic defibrillators, and ED wall clocks. The
Table 1.
Actual data—absolute value of the error of timepieces in minutes before (phase 1), 1 month (phase 2), and 4 months (phase 3) after a one-time community-attempted synchronization event. Timepiece
Phase 1 (No., 95% CI)
Firefighter wristwatches Firefighter AEDs Paramedic wristwatches Paramedic defibrillators Emergency nurse and physician watches ED wall clocks Totals
1.82 (69, 1.25–2.38)* .74 (28, .19–1.29)* 1.20 (42, .79—1.62)* 3.12 (53, 2.44–3.81)* 1.83 (120, 1.51–2.15)* 2.73 (81, 2.15—3.30)* 2.04 (393, 1.82–2.26)*
Phase 2 (No., 95% CI) 1.08 (64, .81–1.35)*† .01 (30, .01–.02)*† .49 (40, .31–.67)*† .84 (51, .26–1.41)*† 1.40 (119, 1.11–1.70)*† .76 (76, .52–1.00)*† .94 (380, .80–1.08)*†
Phase 3 (No., 95% CI) 1.49 (67, 1.03—1.95)† .59 (30, .36–.81)† .78 (33, .44–1.13)† 2.26 (50, 1.60–2.93)† 1.90 (126, 1.51–2.29)† 2.11 (74, 1.45–2.77)† 1.72 (380, 1.49–1.94)†
Results presented as mean number of minutes timepieces deviated from the actual atomic clock time. *P<.05. †P<.05.
APRIL 1998
31:4
ANNALS OF EMERGENCY MEDICINE
4 8 5
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
least improvement was noted in watches worn by firefighters, emergency nurses, and emergency physicians. The error of all timepieces was not statistically significantly different from baseline in any timepiece category 4 months after the one-time attempted synchronization event. Simulation of cardiac arrest cases based on random sampling of the timepieces (Table 2) predicted that, before attempted synchronization, 93% of cases would contain a documentation error of 2 minutes or more and 41% of cases would contain a documentation error of 5 minutes or more. At 1 month, the synchronization event cut the 2-minute documentation error rates almost in half, and reduced the 5-minute documentation error rate by more than three fourths. The error rates nearly returned to baseline by 4 months after the attempted synchronization event. DISCUSSION
Our study confirms that serious documentation errors can result from lack of time synchronization in an urban EMS system. However, unlike previous investigations, we quantified the effect and short-lived benefits of a communityattempted synchronization event. Our results confirm the need for a continuous program of time synchronization in an urban EMS system. Investigators studying similar systems have found that time synchronization is a common problem and the subject has been the focus of editorial commentary.1-5 Cordell et al4 found an average 1-minute 45-second–absolute difference (range of 12 minutes 34 seconds slow to 7 minutes 7 seconds fast) between timepieces sampled on a single day in Indianapolis EDs, ambulances, and fire vehicles and the atomic clock standard. Becker et al2 documented time discrepancies in 39% of prehospital cardiac arrest cases in Chicago. Others have noted significant differences between Table 2.
Simulation data—percent of cases with ≥2- and ≥5-minute documentation errors that would be present on documents recording events on simulated cardiac arrest cases using timepieces with accuracy determined during phases 1, 2, and 3.
No. Cases with ≥2-minute documentation error (%) Cases with ≥5-minute documentation error (%)
4 8 6
Phase 1
Phase 2
Phase 3
2,500 93
2,500 52
2,500 83
41
10
23
paramedic “trip report” times and actual time determined from real-time audio recordings during the runs.1,3 Our investigation included a much larger number and variety of timepieces than those sampled by other investigators. We also explored the implications of inaccurate timepieces by mathematically simulating the cardiac arrest documentation errors that would occur if prehospital cardiac arrest records were generated by personnel using the timepieces to document emergency medical care during resuscitation. Our findings raise serious questions about the medical and medicolegal accuracy of time-based documentation during life-or-death emergency medical treatment. The use of erroneous timepieces can have devastating effects on the documentation of resuscitation emergencies such as those caused by cardiac arrest or major trauma. In these conditions, documentation usually involves multiple timepieces and a large number of events occur per unit of time. This makes it difficult or impossible to reconstruct events accurately for medical, quality improvement, medicolegal, or research purposes. EMS research that is based on the time sequence of events should take into account the special challenges faced by the problem of time synchronization. It seems appropriate for researchers to describe the way in which time is measured in the EMS system, the techniques that are used to standardize timepieces, and the amount of variability that was present among timepieces during the study period. These details should be added to the uniform elements of data collection already recommended for events such as cardiac arrest.5 Ideally, all clocks should be set universally by a central server. For clocks that are controlled by a computer, there are numerous inexpensive programs available on publicly accessible software networks that will command the computer to automatically dial and download the correct time into its timepiece from the atomic clock on a daily basis. Limitations of this study include the fact that only timepieces used to create the records, rather than actual records, were checked for accuracy. In addition, there was no way to “guarantee” that all timepieces were perfectly synchronized immediately after the attempted synchronization event. Another study limitation is that it was conducted in an EMS system that is noted for its excellence in clinical care and continuous quality improvement program. As such, it is likely that the results obtained may actually underestimate the extent and severity of the problem of time synchronization in other, less advanced systems. In addition, the investigation did not sample timepieces in areas of the hospital other than the ED. Thus it would be improper to
ANNALS OF EMERGENCY MEDICINE
31:4
APRIL 1998
SYNCHRONIZATION OF ED TIMEPIECES Ornato et al
extrapolate the findings of this study to other clinical areas. Another potential limitation is that the study focused on the “worst-case” scenario, in which health care workers do not realize that a time documentation error has occurred and do not attempt to adjust for it or account for it in their narrative report. In actuality, some time documentation errors are discovered by health care workers and corrected or discussed in the medical record. Results of this study indicate that timepieces used by EMS providers are often inaccurate, making it difficult to reconstruct events for medical, medicolegal, or research purposes. A community synchronization of timepieces to the atomic clock can reduce the problem significantly, but the effects of a one-time attempted synchronization event are short-lived.
Reprint no. 47/1/88737 Address for reprints: Joseph P Ornato, MD Department of Emergency Medicine Medical College of Virginia 401 North 12th Street, Room G248 Box 980525 Richmond, VA 23298-0525 804-828-7184 Fax 804-828-8597
REFERENCES 1. Mosesso VN Jr: The most neglected tool in EMS: The clock. Ann Emerg Med 1993;22:1311-1312. 2. Becker LB, Ostrander MP, Barrett J, et al: Outcome of CPR in a large metropolitan area— Where are the survivors? Ann Emerg Med 1991;20:356-361. 3. Spaite DW, Valenzuela TD, Meislin HW, et al: Prospective validation of a new model for evaluating emergency medical services systems by in-field observation of specific time intervals in prehospital care. Ann Emerg Med 1993;22:638-645. 4. Cordell WH, Olinger ML, Kozak PA, et al: Does anybody really know what time it is? Does anybody really care? Ann Emerg Med 1994;23:1032-1036. 5. Cummins RO, Chamberlain DA, Abramson NS, et al: Recommended guidelines for uniform reporting of data from out-of-hospital cardiac arrest: The Utstein style. A statement for health professions from a task force of the American Heart Association, the European Resuscitation Council, the Heart and Stroke Foundation of Canada, and the Australian Resuscitation Council. Circulation 1991;84:960-975. 6. Zar JH: Biostatistical Analysis, ed 2. Englewood Cliffs, NJ: Prentice-Hall, 1994:199-201.
APRIL 1998
31:4
ANNALS OF EMERGENCY MEDICINE
4 8 7