Secondary Logo

Journal Logo

Economics, Education, and Policy

Forecasting Preanesthesia Clinic Appointment Duration from the Electronic Medical Record Medication List

Dexter, Franklin MD, PhD*,†; Witkowski, Thomas A. MD; Epstein, Richard H. MD, CPHIMS

Author Information
doi: 10.1213/ANE.0b013e31823fba9e

When scheduling clinic appointments, patient waiting time and staff idle time can be reduced by scheduling patients expected to have different visit durations for correspondingly different lengths of time (e.g., not all for 30 minutes).1,2 There then is an appropriate balance throughout the day between the workload and the providers to care for the patients.35 Accurate estimation of the mean evaluation time reduces patient waiting times, quantified using the minutes late when the patient evaluation starts after the scheduled time or 0 minutes when the patient is seen earlier than scheduled.

Previous studies of anesthesia clinic scheduling performed using simulation and/or queuing theory predicted each patient's appointment duration using the patient's American Society of Anesthesiologists' (ASA) physical status.35 This can be challenging to implement. Electronic medical record entries completed by primary care physicians do not include ASA physical status. Patients do not know their ASA physical status. Assignment of ASA physical status has poor to moderate interrater reliability among anesthesiologists.68

In contrast to the ASA physical status, every electronic medical record contains a medication list. Maintaining an active medication list is a critical (and, in the USA, required) component of the “meaningful use” of electronic medical records. Health systems generally have the medication list updated by a nursing assistant or nurse at the time of each clinic visit (e.g., a surgical consultation or health maintenance encounter). Thus, a recent medication list is available when the preanesthesia clinic appointment is scheduled.

We hypothesized that the count of medications from the medication list would be a more accurate predictor of the time taken by a nurse practitioner to evaluate the patient than the ASA physical status (Fig. 1). If true, then this would provide a scientific foundation for our experience that medication count is a convenient way to choose appointment duration, among health systems with integrated electronic medical records.

Figure 1
Figure 1:
Each increase in the number of medications was associated with longer mean appointment duration (circles) and larger variability in duration. Similarly, longer mean durations were associated with larger variability10 in appointment durations. The variability is represented by the differences between 75th percentiles (top of vertical gray bar) and the mean and between the mean and the 25th percentiles (bottom of vertical gray bar). The data shown are for medications with at least 179 visits. The next largest value (not shown) was 129 visits for patients with 21 medications.

METHODS

The study was approved by the Jefferson Medical College IRB without requirement for written patient consent. From Monday, January 2, 2006, through Friday, May 27, 2011, there were 70,704 visits to the preoperative clinic during which a certified registered nurse practitioner performed the initial evaluation. The interval recorded automatically by the Dräger Innovian (Dräger USA, Telford, PA) preoperative electronic medical record was from the time when the nurse practitioner opened each patient's record until the practitioner saved the evaluation. These times of evaluation were obtainable for 69,654 (98.5%) of the records.a To obtain the best estimates of predictive ability possible, we used all the data available starting 2 weeks after the preanesthesia clinic electronic medical record was put into production.a The study was designed retrospectively (i.e., there was not bias in data collection to support our hypothesis).

We tested whether the number of medications had a larger Pearson correlation with duration than did each of 8 other demographic variables (Tables 1 and 2).9 The Pearson (linear, parametric) correlation coefficient was used because the objective was to forecast time, not to test for rank association. We considered a medication to be whatever any clinician chose to enter into the medication list of the electronic record, including medications taken on an as-needed basis and nonprescription substances such as vitamins or herbs. For each of the nonbinary independent variables X in Table 2, we considered ±X and 12 standard transformations: ±X−2.0, ±X−1.0, ±X−0.5, ±log10(X), ±X+0.5, and ±X+2.0. The transformations were applied so as not to limit the analysis to linear correlations. Our inclusion of the sign (±) simplified presentation in Table 2 and resulted in our effectively comparing the absolute values of the Pearson correlation coefficients. For each variable, we report in Table 2 the Pearson correlation for the transformation (or ±X) giving the largest correlation. For example, the square root of the number of medications was used (Fig. 2). The dependent variable was the log10 (duration) (Table 3), which followed a normal distribution (Fig. 3).3 Because of the large sample sizes and choice of transformations to maximize the Pearson correlation coefficients, all correlations were significantly larger than zero (n = 9 Bonferroni corrected, P < 0.0001), even the r = 0.028 in the last row. We report the n = 8 Bonferroni-corrected, 2-sided comparisons of the Pearson correlation coefficient for medications with that of each of the other 8 demographic variables. All calculations were asymptotic and performed in Excel 2010 (Microsoft, Redmond, WA) using the Fisher transform of the Pearson correlation coefficient.

Table 1
Table 1:
Demographic Information for the Count Data
Table 2
Table 2:
Demographic Information for the Continuous Variables
Table 3
Table 3:
Pearson Correlation Coefficients Between (log10 Durations of Appointments for Preanesthesia Evaluation with Certified Registered Nurse Practitioner) and (Each of Several Predictors from Tables 1 and 2)
Figure 2
Figure 2:
Best linear least-squares fit between log10 duration of visit with certified registered nurse practitioner and the square root of the number of medications. Each circle shows the mean of the log duration for the specified number of medications. The least-squares line was fit with all of the data, not with the means themselves. The 95% confidence bands for the line are not shown because their maximum deviation from the red line is only 0.004 log10 units.
Figure 3
Figure 3:
Histogram of the residuals in log10 units of least-squares linear regression between log10 appointment duration and square root of the number of medications from Figure 2. The superimposed red line is a fitted normal distribution curve. The close qualitative fit shows that assumption of a log-normal distribution3 for analyzing the Pearson correlation coefficients is reasonable given the large sample size (Table 1).

RESULTS

Tables 1 and 2 list the demographic information. The ASA physical status was not known ahead for 82% of patients (i.e., could not be used for forecasting).

The number of medications was a more accurate predictor of appointment duration than any of the other variables (each corrected P < 0.0001), including ASA physical status (Table 3, Fig. 2).

DISCUSSION

Hospitals with enterprise-wide electronic medical records all maintain active medication lists. Scheduling time in the clinic for the preanesthesia evaluation can be done based on the number of different medications the patient was taking at the time of the most recent clinical encounter. Consider schedulers who routinely access the electronic medical record to obtain preauthorizations for tests, determine whether consults have been completed, sequence clinic appointments based on their medical relationships, etc.10 Looking at the patient's medication list can add approximately 10 seconds to the time for scheduling appointments. The corresponding time to plan for the appointment can be obtained from a table (e.g., printed and kept at the scheduler's desk).11

For example, consider patients scheduled for clinic appointments using slots with increments of 5 minutes. If all patients were scheduled for the same amount of time, the duration could be 35 minutes, where 35 minutes ≅ 1.21 × mean of 25.9 minutes (Table 2). The value of 1.21 compensates for factors that tend to increase patient waiting.3 If unequal times were used, then patients with 0 to 2 medications could be scheduled for 25 minutes, where 25 minutes ≅ 1.21 × mean 20 minutes (Fig. 1). Patients with 12 to 20 medications could be scheduled for 45 minutes, where 45 minutes ≅ 1.21 × mean 34.6 minutes (Fig. 1).3

The Pearson correlation coefficient between medications and durations of appointments was 0.340 after transformation (Table 3). The value with ASA physical status was 0.317. Although larger (P < 0.0001), the increased accuracy is too small to be of managerial importance. Number of medications is a better predictor than ASA physical status because medications are known more often prospectively. The 82% incidence of patients without ASA physical status was comparable to that reported elsewhere (Table 1, footnote a).9 The usefulness of our findings may be limited for a clinic with more returning patients, but then the ASA physical status may be less predictive of evaluation time.

Large variance in appointment durations remains after controlling for the numbers of medications (Table 3). Improvement in forecasting accuracy beyond use of medications includes specification of the nurse practitioner evaluating each patient3 (i.e., practitioners significantly differ in their evaluation times [not shown]). A limitation to preanesthesia evaluation clinic management is lack of knowledge of how best to assign patients upon arrival to practitioners.3

RECUSE NOTE

Franklin Dexter is the Statistical Editor and Section Editor for Economics, Education, and Policy for the Journal. This manuscript was handled by Steve Shafer, Editor-in-Chief, and Dr. Dexter was not involved in any way with the editorial process or decision.

DISCLOSURES

Name: Franklin Dexter, MD, PhD.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Franklin Dexter reviewed the analysis of the data and approved the final manuscript.

Name: Thomas A. Witkowski, MD.

Contribution: This author helped write the manuscript.

Attestation: Thomas A. Witkowski approved the final manuscript.

Name: Richard H. Epstein, MD, CPHIMS.

Contribution: This author helped design the study, conduct the study, analyze the data, and write the manuscript.

Attestation: Richard H. Epstein has seen the original study data, reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.

a The timing data were not available from occasional technical access problems with an earlier version. Nurses handled the system being “offline” by manually completing the clinic forms at the time of the visit and entering the data electronically once the system came back online. The consequence of this alternative workflow was that the durations of evaluations logged did not reflect the actual patient encounter times, rather, the post hoc data entry times. Nonetheless, based on the minimum n = 68,080 in Table 3, the largest observed Pearson correlation coefficient of 0.340, and 8 two-sided comparisons, there was >90% power with α = 0.05 to detect a difference for correlation coefficients ≤0.320, which exceeds the second largest observed correlation.
Cited Here

REFERENCES

1. Cayirli T, Veral E, Rosen H. Designing appointment scheduling systems for ambulatory care services. Health Care Manag Sci 2006; 9: 47–58
2. Cayirli T, Veral E, Rosen H. Assessment of patient classification in appointment system design. Prod Oper Manag 2008; 17: 338–53
3. Dexter F. Design of appointment systems for preanesthesia evaluation clinics to minimize patient waiting times: a review of computer simulation and patient survey studies. Anesth Analg 1999; 89: 925–31
4. Edward GM, Das SF, Elkhuizen SG, Bakker PJM, Hontelez JAM, Hollman MW, Preckel B, Lemaire LC. Simulation to analyse planning difficulties at the preoperative assessment clinic. Br J Anaesth 2008; 100: 195–202
5. Zonderland ME, Boer Fredrik, Boucherie RJ, de Roode A, van Kleef JW. Redesign of a university hospital preanesthesia evaluation clinic using a queuing theory approach. Anesth Analg 2009; 109: 1612–21
6. Mak PHK, Campbell RCH, Irwin MG. The ASA physical status classification: inter-observer consistency. Anaesth Intensive Care 2002; 30: 633–40
7. Ragheb J, Malviya S, Burke C, Reynolds P. An assessment of interrater reliability of the ASA physical status classification in pediatric surgical patients. Pediatr Anesth 2006; 16: 928–31
8. Burgoyne LL, Smeltzer MP, Pereiras LA, Norris AL, De Armendi AJ. How well do pediatric anesthesiologists agree when assigning ASA physical status classifications to their patients? Pediatr Anesth 2007; 17: 956–62
9. O'Neill L, Dexter F, Wachtel RE. Should anesthesia groups advocate funding of clinics and scheduling systems to increase operating room workload? Anesthesiology 2009; 111: 1016–24
10. Dexter F, Xiao Y, Dow AJ, Strader MM, Ho D, Wachtel RE. Coordination of appointments for anesthesia care outside of operating rooms using an enterprise-wide scheduling system. Anesth Analg 2007; 105: 1701–10
11. Smallman B, Dexter F. Optimizing the arrival, waiting, and NPO times of children on the day of pediatric endoscopy procedures. Anesth Analg 2010; 110: 879–87
© 2012 International Anesthesia Research Society