| JOURNAL HOME | CME HOME | THIS MONTH | PAST ISSUES | ETOC | COLLECTIONS |
| AUTHORS | REVIEWERS | EDITORIAL BOARD | FEEDBACK | RSS | HELP |
| ||||||||||||||
| ||||||||||||||
|
|
|
|
||||||||||||
|
||||||||||||||
From the Department of Anesthesia and Critical Care, Massachusetts General Hospital, Boston, Massachusetts.
Address correspondence to Robert A. Peterfreund, MD, PhD, Department of Anesthesia and Critical Care, Jackson 439, Massachusetts General Hospital, 55 Fruit Street, Boston, MA 02114. Address e-mail to rpeterfreund{at}partners.org.
 |
Abstract |
|---|
 Top Abstract IntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
METHODS: We reviewed all electronic anesthesia records generated during a 1-month period at our institution to ascertain completion rates for six clinical documentation elements: allergies, IV access, electrocardiogram rhythm, ease of mask ventilation, laryngoscopic grade of view, and insertion depth of the endotracheal tube.
RESULTS: Of 2838 records, 64% had the necessary free text remark in the allergy element. The free text required to complete endotracheal tube depth documentation appeared in 538 of 918 cases in which the patient was tracheally intubated (59%). Free text documentation of the electrocardiogram rhythm diagnosis appeared at least once in 86% of records. Documentation of mask ventilation characteristics was entered by touch screen from a pick list and was expected in 781 records but appeared in 664 records (85%). Laryngoscopic grade of view documentation was also selected by touch screen and expected in 883 records but present in 811 cases (92%). Any notation of IV access appeared in 84% of records.
CONCLUSIONS: We found that electronic clinical anesthesia documentation was often incomplete. Dependence on free text remarks and the record keeping system’s inability to automatically present entries in logical sequences consistent with workflow were associated with incomplete data entry. Our results suggest that the user interface for data entry, and the logic that an electronic system uses for preventing omissions and inconsistencies, merit further study and development in order to facilitate clinically useful documentation.
|
Introduction |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
An AIMS generates an anesthesia record using a combination of data collection methods. Some demographic, patient identifier, and scheduling information can be captured automatically via interfaces with central hospital data systems. Other interfaces automatically capture patient physiologic monitoring data, such as arterial blood pressure, temperature, heart rate, hemoglobin oxygen saturation, and end-tidal gas concentrations, from medical devices directly linked to the computerized system. Completion rates for such automatically captured entries should be 100%.
The AIMS also captures data entered directly by the clinician. Depending on the user interface design, the clinician can 1) select via a touch screen or mouse click preconfigured objects or elements built into the application (e.g., drugs, fluids, or gases administered; coded entries such as the times of anesthesia induction or start of surgery), or, 2) type free text into remarks fields using a keyboard to complete a data element. To investigate the completeness of clinical documentation when using an AIMS, we reviewed all records generated during a 1-month period by the AIMS used at our institution. We tested the hypothesis that features of the application and data entry methods influence rates of completion for basic clinical elements of an anesthesia record.
The entry methods compared were free text entry and touch screen selection from a menu or pick list. The data elements for allergies, electrocardiogram (ECG) rhythm, and depth of endotracheal tube (ETT) insertion require free text entry. Completion of the data elements for ease of mask ventilation and grade of laryngoscopic view is accomplished by selecting options from a pick list displayed on the touch screen. With the AIMS as configured at our institution, IV access documentation can be completed either by free text entry or by mouse click from a pick list. We measured the entry of these clinical data elements to evaluate whether the entry mechanism affects completeness of clinical documentation.
|
METHODS |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
System Development and Customization
The Saturn AIMS provides a structured database, various interfaces to populate data from external sources such as medical devices, and application software for the user interface. Screen layouts have tabs for data entry at each phase of anesthesia care. The end user can customize portions of the native application. Customized data elements can be constructed as "events," which are added to the intrinsic features of the native application. One property of events is that they appear in the final printed record. We customized touch screen selectable content in the form of pick lists for certain data sets within folders or screens. Examples of selectable content include lists of IV site and size combinations (Fig. 1, right), drugs, fluids, laboratory studies, and clinical occurrences or clinical notes. We minimized reliance on free text entry because it becomes difficult to organize and query the resulting data, negating an advantage of an AIMS.
|
The entries created for clinical elements were set-up according to a coding system created to be intuitive and allow for organizing and sorting data elements within the application. For example, the event "L200R IV: 20 gauge IV, right hand" can be selected from a lengthy catalogue of multiple possible IV catheter sizes and insertion sites (Fig. 1, right). After populating the system with the possible entries, we preconfigured templates or "environments." These are packaged combinations of routine entries including drugs, fluids, and clinical occurrences based on service- or procedure-specific chronological order. All templates shared entries for basic elements of an anesthesia record, such as allergies, ECG, and IV access data. The combinations in these templates are contingent on the various anesthesia methods and activities for a particular case. This facilitated data entry for various anesthesia techniques and types of cases, such as routine general anesthetics, craniotomies, regional block anesthetics, cardiopulmonary bypass cases, etc. Cues from the intuitive alphanumeric codes associated with the specific elements could then be used to find similar types of potential entries in the body of the whole catalog that are not part of the particular environment template. Owing to limitations of the application and the layout of the AIMS-generated printout, unique conventions were created to fulfill our institution’s documentation requirements for components of an anesthesia record such as ECG rhythm diagnosis and allergy information. Customization enabled us to use the Saturn AIMS to generate a printed anesthesia record resembling our traditional handwritten record.
AIMS Workflow
When creating the anesthesia record using the AIMS, providers select the appropriate patient from a schedule of cases, and then load templates intended to facilitate documentation for the specific type of case. The application automatically records in real-time physiologic data generated by monitors interfaced with the system. The clinician enters other data via the user interface. As patient care necessarily takes precedence over documentation, some user-entered data are entered retrospectively. The user can adjust the timestamps for such entries to approximate the actual time of occurrence. Clinicians can choose to enter data such as coded events and drug entries directly from a template, add free text to these entries, or search the whole catalog of drugs and events to select additional appropriate entries if the desired entry is not part of the loaded template. Other entries can be made directly from the chart layout screen generated by the environment. The user interface of the main Intra-Op charting tab of the application is shown in Figure 2.
|
Once the case ends and anesthesia data recording terminates, the clinician finalizes the anesthesia record by printing a paper document. The printout is placed in the patient chart in the recovery location.
Study Period and Data Points
We retrospectively examined all 2838 cases stored in the AIMS during January, 2005. This interval was selected to ensure that clinicians entering training the previous summer were facile with AIMS charting. We selected data entries that would be expected to appear in a complete anesthesia record to analyze for the completeness of data entry. These data entries included patient allergies, IV access, ECG rhythm, ease of mask ventilation, grade of laryngoscopic view, and depth of ETT insertion. These are documented in the AIMS, as configured at our institution, either by free text entry or by touch selection from a pick list.
Data Collection
We queried the AIMS completed case database using Crystal Reports (Crystal Reports version 8.5 software, Business Objects Inc., San Jose, CA) to extract targeted data meeting specific, predetermined, filtering criteria. To electronically identify records of cases with insertion of an ETT for airway management, we used documentation of cisatracurium administration, the predominant nondepolarizing muscle relaxant used in our department, as the filtering criterion. Only records from patients admitted to the hospital on the day of surgery, or patients undergoing their procedures in the Surgical Day Care Unit, were included in the analysis. This excludes from the analysis patients who already had an ETT before coming to the operating room (e.g., patients from the intensive care unit), but may have included in the analysis a very small number of patients with chronic tracheostomies. We maintained anonymity by extracting specific targeted information and organizing reports by system-generated case identification numbers. Also, to eliminate subjective interpretation of the clinical benefit of the recorded data, the occurrence of any documentation for the targeted data elements passing the filters was included in analyses. We exported the reports to a spreadsheet for pooling and quantitative analysis.
|
RESULTS |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
|
Bag-Mask Ventilation
In the general endotracheal anesthesia template, clinicians can select by touch screen one of three coded events to qualitatively document the ease of mask ventilation (A101 Mask ventilation easy, A102 Mask ventilation fair, or A103 Mask ventilation difficult). The routine template does not include a choice for failed or impossible mask ventilation. We examined completion of documentation characterizing the ease of mask ventilation before tracheal intubation. Refining the query to select records documenting cisatracurium as the only muscle relaxant administered for the case excluded records in which succinylcholine appears. The assumption was that cases with rapid sequence inductions, during which mask airway characteristics are not assessed, would be excluded from the data set. Applying this selection criterion reduced the total number of cases qualifying for analysis to 781. An occurrence of any one of the three data elements documenting mask airway characteristics appeared in 664 of the 781-targeted cases (85% of the records). We anticipate that a small error was introduced by our filtering criteria since a few patients may have received succinylcholine to facilitate endotracheal intubation even when rapid sequence techniques were not used, or when cisatracurium (typically in high doses) was used for rapid sequence induction (no mask ventilation) in situations in which the depolarizing relaxant was contraindicated.
Grade of View During Laryngoscopy
In the template for general endotracheal anesthetics, clinicians can select by touch screen one of four coded events to document the grade of the view during laryngoscopy (A301 Laryngoscopy: Grade 1, vocal cords seen; A302 Laryngoscopy: Grade 2, arytenoids seen; A303 Laryngoscopy, Grade 3, only epiglottis seen; A304 Laryngoscopy, Grade 4, epiglottis and larynx not seen). We found that in 811 of the 883 records (92%) in this targeted data set, clinicians selected one of the four coded events for laryngoscopic grade of view. A search of the remaining records revealed no documentation of failed intubation, emergency surgical airway, or case cancellation attributable to inability to intubate the trachea.
Data Entry by Free Text Entry
Patient Allergy Data
The native Saturn application provides an allergy module only within the preoperative screen. Allergy data entered in this module print only on the preoperative report, a feature not in use at our institution, instead of on the final intraoperative anesthesia record. To ensure that allergy data appeared on the final anesthesia record, as is standard at our institution, a customization created an event for allergy documentation, because events do appear on the anesthesia record printout. We custom configured the event "1210 Allergies: ______" to appear automatically as the first note on the anesthesia record as each record is started. The clinician is expected to type free text in the remark section (Fig. 3) to complete the entry. Of 2838 records, 1806 (64%) had a text entry of allergy descriptors (either no known drug allergies or the report of a specific drug allergy and reaction) in the allergy remarks field. The remarks field in all other records remained blank.
|
Insertion Depth of an ETT
An element based on the prerequisite of ETT insertion during the anesthetic is documentation of the insertion depth of the tube. Of the 918 cases in which an ETT was placed, 721 records (79%) contained the coded event, "A490 ETT: secured at (cm): ______." This coded event is selected by touch screen. Selection of the event is insufficient to complete data entry. The user must enter free text to document the tube insertion depth in the remarks field of this data element. The required free text appeared in 538 records, 59% of the 918. This also reveals that 183 of 918 records (20%) contained an event stating the tube was secured at a certain depth, but no free text note was entered to document that depth.
ECG Rhythm
The monitoring systems available in our operating rooms transmit to the AIMS heart rate data from the ECG. The ECG rhythm is not automatically diagnosed or reported. The anesthesia provider must make this diagnosis and is expected to document it at regular intervals on the anesthesia record. Our traditional paper anesthesia record has a grid row to log the ECG rhythm diagnosis at 15 min intervals. Our native AIMS application does not provide a mechanism to readily record this information. We therefore generated another customization using the application’s laboratory data entry feature to facilitate this entry (Fig. 4), so that the AIMS-generated final printout would include these data in a familiar manner. With this customization, completion of the entry for ECG rhythm allows text entry directly into the chart grid on the main Intra-Op screen of the application.
|
Of the 2838 records reviewed in this study, 2431 (86%) had at least one free text entry documenting the ECG rhythm diagnosis sometime during the case. The ECG rhythm diagnosis appears as a row entry on the electronic chart screen as well as on the printed record in timestamped columns (Fig. 4). If no entry is made in the application during the case, then the row for ECG rhythm does not appear on the final printed anesthesia record. We noted that in many anesthesia records, the ECG diagnosis is recorded only once, or sporadically throughout the case, rather than at regular intervals (Driscoll et al., in preparation). The criterion used to identify completion of ECG data entry in this study, the occurrence of at least one documentation of the ECG diagnosis by free text entry, thus overestimates the true completion rate for this clinical variable.
Data Entry Either by Touch Screen Selection or by Free Text Entry
Initially, the Saturn AIMS was custom-configured to contain a lengthy, intuitively coded, pick list of various common combinations of IV catheters and insertion sites. Shortly after the introduction of the AIMS into routine clinical use, anesthesia providers reported that scrolling through the long list of specific choices was slow and inefficient. Consequently, two generic events were created, "L105 IV site: ______" or "L106 IV size (ga): ______" (Fig. 1, left). These are included in templates and selectable by touch screen. However, each entry requires free text to complete the documentation of information. Consequently, clinicians have two distinct options for entering data about IV placement.
At least one IV access event of any kind appeared in 2384 (84%) of 2838 records. Of the 2384 records in the study period containing any IV access documentation, the generic event (either L105 or L106, or both together) from the template was used 80% of the time. However, in approximately 20% of the cases in which a generic IV access event was selected, there was no free text description added to the remarks field (375 of 1858 occurrences of event "L105 IV site:;" 345 of 1590 occurrences of event "L106 IV size (ga):"). Specific events were selected from the catalog to complete the IV access documentation in 471 (17%) records. Thus, when given an option, clinicians preferentially chose the generic documentation entry, even though the entry was not completed in 20% of those selections. In the remaining 3% of the cases (117 records), specific IV access events appear automatically as part of the template. An example is the template for outpatient electroconvulsive therapy procedures, in which every patient receives a 22-gauge IV. Data are summarized in Table 1.
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
To produce an anesthesia record meeting our institutional requirements, we adapted some of our workflow and documentation requirements to the application’s capabilities. We also used the customizable features, either as they were intended or in an unconventional manner, to facilitate capture and presentation of the data elements on the final AIMS printout that appear in our traditional handwritten record.
A comparison of the two data entry methods, touch screen selection of a specific data element and free text entry into a remarks field, suggests that information entry is less likely to be complete if the clinician must generate free text. Thus, we found low completion rates for text entry into the allergy remarks field (64%) and the depth of insertion of the ETT remarks field (59%). Completion rates for touch screen selected specific data elements were higher, 85% for mask airway characterization and 92% for grade of view of the glottis during laryngoscopy.
We speculate that clinicians may leave the allergy remark field blank as a default if there are no known drug allergies. Workload and time pressure may detract from attention to completing the depth of insertion data element that requires free text entry. The allergies data element as presently designed either in the native application or in our customized method provides no additional clinical benefit because this AIMS does not confirm allergy information with another source and does not check for administered drugs against the allergy data. We speculate that allergy documentation would be greatly improved if data were automatically linked to other hospital systems.
A method to ensure entry of allergy or other essential data would be creating a "hard stop" so that a record could not be closed if the data element remained incomplete at the conclusion of a case. A disadvantage is the amount of programming and number of warning messages necessary to enforce completion of the many possible required elements of a record. Alternatively, if clinicians are prompted to enter an item of data, such as allergy information, and the data entry point is easily accessed, they may be more likely to comply. The number of prompts, however, may be significant and their appearance may be distracting during clinical care.
Under the filtering conditions of this study, which screened for any entry into an ECG Rhythm grid cell at any point during a case, we observed an 86% completion rate without prompting. We suggest that clinicians generally accept the method of ECG data entry into a customized laboratory result feature (Fig. 4) due to its relative ease and familiarity, and the default appearance on the grid. However, completion of this data element appears to be inconsistent throughout a case (Fig. 4, and Driscoll et al., in preparation). Consequently, the data likely overestimate the true completion rate, if entry of the ECG diagnosis consistently throughout the entire case is defined as the requirement for this clinical element of the record. We speculate that reduced compliance in documenting the ECG rhythm diagnosis would have occurred if we had chosen another data entry method. Compliance with data entry throughout a case might improve if the AIMS could prompt clinicians to periodically document ECG rhythms. However, the particular commercial platform’s software does not allow required laboratory results, which was the data entry route for ECG rhythm. A prompting function is thus not available for data entered by this modality.
We initially established a lengthy, systematic, custom, subcatalog of every possible combination of IV sizes and insertion sites. During pilot use of the AIMS, we observed that clinicians rarely selected specific coded events for IV insertion site and size. Therefore, IV access documentation was omitted or documentation was inaccurate. We revised our IV documentation workflow by adding the touch screen selectable generic IV site and size events to intraoperative templates (Fig. 1). The data reveal that clinicians are more likely to choose an event from a template (83%) rather than search through the catalog of events (17%). However, the generic events require free text entry to complete the entry. The remarks fields into which the clinician would enter such information were often left blank. As is the case for allergies or ETT insertion depth, IV access documentation might be improved by prompts to enter IV documentation or by making IV documentation a required field in the record.
The data elements of either a handwritten record or an AIMS-generated record can be grouped into two broad conceptual categories. Some data are explicitly required; others are conditionally required. Documentation of explicitly required data would be expected for all cases. Examples of explicitly required data include administrative information such as patient demographics, case start times and end times, compliance with regulatory requirements, and fundamental patient physiologic variables. Depending on institutional practice, other explicitly required elements might include allergies, ASA physical status classification, documentation of initial IV access, and documentation of monitoring.
Conditionally required components of the anesthesia record are entered only when certain sets of criteria are met, whether planned or unplanned, depending on the course of events during a case. Examples of conditionally required components include details about the intubation procedure if a general endotracheal anesthetic is administered, or details describing the placement of a regional anesthetic if a nerve block is chosen for the primary anesthetic. To ensure that conditionally required data elements, such as a description of the ability to mask-ventilate a patient, are completed in a record documenting a routine general endotracheal anesthetic, we propose that an ideal AIMS should have the ability to detect the absence of essential information. The AIMS should then facilitate and mandate completion of data entry.
At our institution, clinicians are expected to document their characterization of the grade of laryngoscopic view. For this documentation requirement, the customized AIMS in use at the Massachusetts General Hospital provides an efficient method (one touch to select an entry from an easily accessible touch screen menu) for completing an essential element of the anesthesia record under the condition of an endotrachael intubation. We hypothesized that if an entry were easily accessible from a template, or in plain view on the screen, completion rates would be high without additional prompting. Our observed completion rate of 92% for documenting laryngoscopic view grade supports this hypothesis. The findings suggest that when free text is required for completion of a selected coded event, the completion rate will be low compared to events requiring selection only. The disadvantage of adding many coded events for selection from a pick list is the resulting length of the pick list and difficulty finding individual items. These observations support the concept that event selections must be easily located, and that a minimum number of steps should be necessary to complete data entry (13).
Overall, we found that incomplete data entry was associated with relying on typed entries to populate coded event remark fields. Incomplete data entry also correlates with the limitations of the application to prompt for, or display, the next logical entry as a means of encouraging complete documentation. Failure of the system to automatically present coded events in logical sequences consistent with workflow also appears to hamper complete data entry. As exemplified by the Saturn AIMS, the limitations of current systems are exposed when they rely on free text entries (inconsistent terminology, difficult to search) or it becomes time-consuming and difficult to enter coded events (searchable catalogs). To achieve the intended benefits of an AIMS, the data entry interface must be well organized, consistent, and simple.
The major limitation of this study is that we analyzed the performance of only one commercially available AIMS. Thus, our findings may not be applicable to all AIMS and all institutions. The Saturn AIMS provides mechanisms for guiding the documentation of various events through templates, and also provides reminders for some required fields. We found that it does not prevent incomplete or clinically inconsistent documentation. Specifically, when there are inconsistent data entry methods, multiple steps to enter data, entries deeply nested in menus, or screens full of unused fields, clinicians are less likely to completely or accurately enter the desired data.
A second limitation of the study is the assumption that the clinicians using the system are equally facile, so that individual performance does not influence the data quality in the record keeping. However, a large number of records were studied and the distribution was over the entire group of clinically active personnel at a time when all clinicians had ample exposure to the application to not be considered novice. Thus, any outlying performance by individuals is unlikely to significantly alter the findings.
Our department mandates completion of particular regulatory, billing, electronic signature, and government compliance data elements for administrative purposes. Onscreen prompts and a variety of follow-up methods including pager and email messages ensure completion of those administrative elements (16). As a result, more than 99.9% of anesthesia records in fiscal 2005 prepared with the Saturn AIMS meet documentation requirements for billing purposes (seven cases unbillable of 35,612 electronic records; Spring and Moss, personal communication). In agreement with other findings (13), our results suggest that the user interface for guiding the clinical data element entry into an AIMS, and the logic and methods these systems use for preventing omissions of clinical data and inconsistencies, can be improved and merit further study.
On the basis of our data, we conclude that data entry of clinical documentation using predefined pick lists results in more complete documentation than when free text entry is used. Even predefined pick lists do not result in 100% inclusion of required documentation. If we are to achieve the goal of 100% complete clinical documentation, future generations of AIMS should be designed based upon studies of the impact of data entry methods on the completeness of documentation. The method of data entry should be evaluated, along with the impact of designs that include required fields and automated reminders, in the context of linking documentation to the dynamic workflow of the anesthesia provider. Additional studies are required to assess how well these systems are working in clinical practice and provide quantitative data to assist software development.
|
ACKNOWLEDGMENTS |
|---|
|
Footnotes |
|---|
Supported by the Department of Anesthesia and Critical Care, Massachusetts General Hospital.
Parts of this work were presented in abstract form at the American Society of Anesthesiologists Annual meeting in Atlanta, GA, on October 26, 2005.
Reprints will not be available from the author.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMETHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  W. S. Sandberg, E. H. Sandberg, A. R. Seim, S. Anupama, J. M. Ehrenfeld, S. F. Spring, and J. L. Walsh Real-Time Checking of Electronic Anesthesia Records for Documentation Errors and Automatically Text Messaging Clinicians Improves Quality of Documentation Anesth. Analg., January 1, 2008; 106(1): 192 - 201. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
