Use of a Time-Flow Study to Improve Patient Waiting Times at an Inner-city Academic Pediatric Practice | Nursing | JAMA Pediatrics | JAMA Network
[Skip to Navigation]
Sign In
December 2002

Use of a Time-Flow Study to Improve Patient Waiting Times at an Inner-city Academic Pediatric Practice

Author Affiliations

From the Division of General Pediatrics, Albert Einstein College of Medicine/The Children's Hospital at Montefiore, Bronx, NY.

Arch Pediatr Adolesc Med. 2002;156(12):1203-1209. doi:10.1001/archpedi.156.12.1203

Objectives  To use findings from a time-flow study at an academic pediatric practice to restructure practice patterns and to repeat the time-flow study to evaluate consequent changes in patient waiting times, total visit times, and room and nursing time usage rates.

Design  A before-and-after time-flow evaluation of patient waiting times, total visit times, and room and nursing time usage rates.

Setting  An inner-city academic pediatric practice located at a community health center affiliated with a major urban academic teaching hospital.

Patients  All patients visiting the pediatric practice during the weeks of April 26 through 30, 1999, and March 24 through 28, 2000.

Interventions  Initial time-flow findings generated in telephone messaging, provider scheduling, nursing location, and provider and preceptor documentation.

Main Outcome Measures  Patient waiting times, total visit times, room usage rates, and nursing time usage rates.

Results  After the implementing of practice changes, mean total visit time declined from 91.9 to 78.3 minutes. Mean waiting time to be called by the nurse decreased 3.2 minutes, mean time spent with the provider decreased 4.6 minutes, and mean time to wait for a preceptor decreased by 8.8 minutes. Multivariate analysis controlling for visit type, provider type, and the type of postvisit interventions found that mean time of visit decreased by 13.6 minutes. Room and nursing time usage rates, resident satisfaction, and quality-of-care indicators remained largely unchanged.

Conclusions  Time-flow studies can be useful instruments for academic ambulatory practices to identify and ameliorate practice inefficiencies without sacrificing quality of teaching or patient care.

ACADEMIC MEDICAL centers have faced recent well-documented threats to financial stability.1,2 While costs continue to rise,3 prospective reimbursement arrangements, greater managed care market penetration, decreases in public health insurance funding, demographic changes, and increasing numbers of uninsured patients have all contributed to limiting reimbursement.4-6 In response to these developments, clinical programs at academic medical centers have attempted to increase the volume of patients seen to offset declines in per-visit revenue. In a separate but contemporaneous trend, residency programs in primary care disciplines such as pediatrics, family medicine, and internal medicine have placed increased emphasis on ambulatory aspects of training.7,8 The combined effect of these 2 coincident trends is that more resident time is being spent in outpatient settings at the very time that these venues are being pressed for increased productivity in the form of larger visit volumes.

How to resolve the competing demands on clinical time of simultaneously increased training and increased visit volume is an important management challenge. Hours can be extended, ancillary staffs augmented, and faculties expanded, but each of these responses generates increased costs. Ratios of trainees to supervising attending physicians can be allowed to increase or the time spent with patients can be curtailed if one is willing to tolerate disaffection among either of these 2 critical constituencies. Yet often overlooked inefficiencies in many academic outpatient clinical settings also provide opportunities to enhance productivity without sacrificing training effectiveness or patient care goals. To identify and correct existing inefficiencies requires a systematic approach and coordination among all aspects of the outpatient site.

We report herein the use of sequential time-flow studies in a large urban pediatric academic practice during two 1-week periods 11 months apart to successfully identify potential areas of improvement in clinical activity. We also report the results of a coordinated effort to exploit these findings by reconfiguring the training and clinical activities at the practice to reduce visit time while maintaining resident satisfaction and improving quality of care.



The study was conducted at an urban academic pediatric practice located in a federally sponsored community health center in the east Bronx, NY, associated with a major academic medical center. The pediatric service at this site is open from 9 AM to 7 PM 4 days a week and from 9 AM to 5 PM 1 day per week. It has 24-hour telephone access for patients and serves an ethnically diverse community of lower-income families from the surrounding community. Approximately 10% of the pediatric population have commercial insurance, 10% are uninsured, and the remainder have publicly financed health insurance.

At the time of the study, the practice employed 11 full-time equivalent (FTE) pediatricians and 0.5 FTE pediatric nurse practitioner. The pediatric service also served as the site at which 29 pediatric house officers saw patients in their weekly continuity practices and as part of their outpatient department rotations. During sessions when resident physicians saw patients, the ratio of resident physicians to attending pediatrician preceptors varied between 2:1 and 3:1 each year of the study. There were 8 postgraduate level 1 trainees (PL-1s), 9 PL-2s, and 5 PL-3s seeing patients during the week studied in 1999 and 9 PL-1s, 9 PL-2s, and 5 PL-3s seeing patients during the week studied in 2000. The nursing staff consisted of 1 registered nurse, 5 licensed practical nurses, and 3 patient care technicians.

We gathered data on every patient visit from Monday, April 26, through Friday, April 30, 1999, and from Monday, March 24, through Friday, March 28, 2000. The volume of visits at the time of this study was approximately 20 000 visits per annum. Visits are categorized as either emergency visits or routine visits. Emergency visits are sick visits for specific acute complaints, whereas routine visits are either for health care maintenance or for health care maintenance follow-up of an ongoing chronic condition. Seasonal variation during the course of the year leads to relatively greater numbers of emergency visits in the winter and early spring and fewer such visits in the summer and early fall. By scheduling the study to occur in late spring for both years, we attempted to hold constant across the 2 study periods any seasonal variation in visit volumes. In addition, we anticipated that the numbers of emergency visits would be somewhat less than at peak times during the year, but not as few as during trough times. During the weeks of the study, there were 426 recorded patient visits in 1999 and 418 in 2000.

Patient flow

In preparation for the study, we outlined the flow of patients through the practice diagramatically (Figure 1). As can be seen in this depiction, patients arrive at the practice and, after notifying the receptionist, are seated in the waiting room until they are called by a nurse to an examination room. All patients, including those with emergency visits, have appointments with a specific provider at a scheduled time. If a patient appears without an appointment, he or she is given a scheduled appointment for later in the session if one is available or for the following day if the schedule is full the day he or she first appears. The nurse then takes vital signs and records the reason for the visit before placing the chart in the consultation room. From there, providers take patient charts into the examination rooms to begin their visits with their assigned patients.

Figure 1. 
Patient flow through the visit.

Patient flow through the visit.

The visits are conducted either by attending physicians or by resident physicians. At an attending physician visit, once the physician is finished, he or she can either discharge the patient directly or place the patient's chart in a rack for the nurses to pick up if immunizations, laboratory tests, or hearing or vision screenings are necessary. Resident physicians must, after completing their visits, present their findings to an attending physician, who discusses the case with them. In accordance with federal regulations regarding resident supervision, depending on the nature of the visit and the level of training of the house officer, the attending physician then proceeds either to see the patient or to place a written indication in the chart of agreement with the management and plan outlined by the resident. Once the precepting is finished, as with attending physician visits, the patients are either discharged directly or transferred to the nurses for postvisit procedures such as immunizations, hearing and vision screening, blood drawing, etc.

Initial data collection, year 1

With the flow of patients outlined diagrammatically, a collaborative effort was undertaken in April 1999 to collect the first week of data on time spent in the practice. Rather than employ a separate individual or team of individuals to track patients through the service, every clerk, nurse, and practitioner recorded the beginning and end of his or her respective intervention with a patient at each successive stage of the visit. This permitted monitoring of every visit during the week and engagement of the entire staff in the process. Each examination room and nursing station contains a telephone that displays a clock time that is standardized throughout the practice. By relying on these clocks, we achieved time synchronization without undue effort.

We created a recording sheet on which each participant entered the time he or she began and finished each portion of the patient's visit (Figure 2). For example, the receptionists recorded patients' arrival times and the times of their appointments. Nurses recorded when they called patients into rooms, when they began and ended their triage, and when they retrieved charts from providers to begin postvisit procedures. Each provider recorded when they began and ended their visits. Intervals between the end of one patient interaction and the beginning of the next were calculated to estimate the time patients waited between different stages of the visit. In this way, every waiting period and every interaction were assigned a time interval.

Figure 2. 
Recording sheet. EV indicates emergency visit.

Recording sheet. EV indicates emergency visit.

The beginning of the visit was timed differently depending on the arrival time of the patient relative to when the visit was scheduled. For patients who arrived on time or early, waiting time to be called by the nurse was measured from the scheduled appointment time, while for patients who arrived late, this interval was measured from the patient's arrival time. For patients who arrived early and were called before their scheduled appointment time, waiting time was coded as 0 minutes. The total time a patient spent in the practice was estimated by subtracting the appointment time from the discharge time for patients who arrived on time or early and was estimated by subtracting the arrival time from the discharge time for patients who arrived late. For patients who were called before their appointment time, total visit time was estimated as the difference between the time they were called by the nurse and the time they were discharged from the visit. The distributions of arrival times relative to appointment times for both years are depicted in Figure 3. In each year, for more than 75% of all visits, the patients arrived within 20 minutes of their scheduled appointment.

Figure 3. 
Arrival times relative to scheduled appointment times, 1999 and 2000.

Arrival times relative to scheduled appointment times, 1999 and 2000.

Usage calculations

Given recorded waiting times, the number of examination rooms available per session, and the nursing FTEs present per session, we were able to calculate the following usage rates:




where UR indicates room usage rate; P, total number of patients seen that session; TR, total mean time spent in room (total visit time − time waiting to be called − [postvisit time × 0.5]) in hours; R, total number of available examination rooms that session; HR, total hours rooms are available; UN, nursing time usage rate; TN, mean nursing time per patient (triage time + postvisit processing time); and HN, total number of nursing hours available for that session (total nursing FTEs available for that session × number of hours worked per FTE).

Resident satisfaction

Twice a year, each resident is asked to anonymously complete a survey that monitors trainees' satisfiaction with a variety of characteristics of the practice, including the magnitude of patient visit volumes, interactions with the faculty, nursing activities, and general efficiency of the practice. We report herein the findings with respect to perceived availability of faculty preceptors before and after the implementation of changes designed to decrease patient waiting times and increase practice efficiency.

Quality of care

Ongoing quality improvement activities at the practice include the use of an instrument to monitor rates of compliance with a comprehensive array of indicators including immunization rates, screening for hemoglobin and lead levels at appropriate ages, evidence of counseling with respect to nutrition, anticipatory guidance and smoking prevention, and other measurable outcomes. We report herein changes in the rates of compliance with these indicators between the first and second time-flow studies as measures of whether quality of care delivered at the practice was compromised by attempts to decrease patient waiting time and increase practice efficiency.

Statistical analysis

Data were entered into an Excel (Microsoft Corp, Redmond, Wash) spreadsheet and then analyzed with Stata software (Stata Corp, College Station, Tex). The t tests were used to detect differences in means for continuous variables, and χ2 tests were used to analyze differences in categorical variables. Wilcoxon–Mann-Whitney rank sum tests were used for nonparametric estimations. Multivariate ordinary least-squares regression analysis was used to estimate coefficients for the amount of visit time accounted for by specific characteristics of the visit. By including year in these regressions, we could determine how much total visit time decreased from year 1 to year 2 holding other visit characteristics constant.


Table 1 documents the distribution of visits in 1999 and 2000 with respect to days of the week, type of visit (emergency visit or routine visit), and type of provider (resident or attending physician). During the first year, a relative imbalance in patient visits resulted in Tuesdays being generally busy and Fridays being generally slow. When the sample of visits studied in 1999 was compared with that of the sample studied in 2000, the proportion of total visits seen by residents increased as the panel size of these practitioners expanded. The mean ratio of residents to preceptors during the week studied in 1999 was 2.4:1, while in 2000 during the week studied the mean ratio was 2.3:1. The proportion of emergency to regular visits did not change significantly.

Table 1. 
Distribution of Visits by Day of Week, Type of Visit, and Type of Provider, 1999 and 2000*
Distribution of Visits by Day of Week, Type of Visit, and Type of Provider, 1999 and 2000*

Usage rates are given in Table 2 for years 1999 and 2000. In general, neither rooms nor nursing times represented capacity constraints in any session. With respect to room usage rates, 6 of the 10 sessions during the week showed decreases in room usage rates from 1999 to 2000 despite an equivalent number of patients in each year. Of the other 4 sessions, 2 (Thursday and Friday afternoons) had room usage rates that were essentially unchanged and 2 particularly underused sessions (Wednesday and Friday mornings) had increases in their room usage rates. Nursing time usage rates similarly declined in 5 of 10 sessions, remained unchanged in 3, and increased in 2 sessions from 1999 to 2000. Neither of these sets of changes reached statistical significance.

Table 2. 
Room and Nursing Usage Rates by Day of the Week and Session, 1999 and 2000*
Room and Nursing Usage Rates by Day of the Week and Session, 1999 and 2000*

Organizational response

An analysis of data from the initial flow study, interviews with staff, and calculations of the nursing and room usage rates disclosed several areas in which potential improvements were deemed possible. These included the following: (1) nursing time was being absorbed by answering telephone callers with essentially administrative questions; (2) the uneven distribution of practitioners throughout the week resulted in overcrowding of examination rooms on some days and their relative underuse on other days; (3) the centralization of nursing functions in 2 nursing stations impeded the ability of nurses to monitor the flow of patients throughout the practice and to respond to patient flow needs as they arose with individual practitioners; and (4) time spent documenting patient visits and precepting appeared to be excessive.

After reviewing these results with physicians, nurses, administrators, and clerical staff, several changes were instituted: the installation of a telephone tree to reroute administrative calls away from nurses' stations and to clerical staff; the reassignment of practitioners to a more even distribution throughout the week; the decentralization of nursing function into 4 pods located throughout the practice; the redesign of age-specific patient visit sheets to include more provider prompts and checklists; and the adoption of precepting stickers that streamlined documentation for attending pediatricians. The numbers of nurses, attending pediatricians, and resident trainees remained unchanged from 1999 to 2000.

Repeating the flow study in 2000 identified changes in time intervals that occurred at different points during the visit. Table 3 records these changes. The first column indicates time intervals and total visit time for 1999. The mean total visit time that year was 91.9 minutes. Of that total time, more than 20 minutes were spent waiting to be called into a room after arrival at the practice. In addition, more than 18 minutes was spent waiting for a provider to begin the visit after the patient went through triage, and an additional 14 minutes were spent waiting for a preceptor to arrive after a resident visit. This table also shows that 22% of the visits needed immunizations, 23% needed laboratory work, and 13% required hearing and vision screenings. The average patient arrived on time for his or her appointment in both years of the study.

Table 3. 
Changes in Time Intervals of Pediatric Office Visits, 1999 and 2000*
Changes in Time Intervals of Pediatric Office Visits, 1999 and 2000*

The second column in Table 3 shows the times recorded the year after organizational changes were instituted at the practice. Mean total visit time decreased 13.6 minutes, from 91.9 to 78.3 minutes. Accounting for the largest portion of that decrease, the mean time to wait for a preceptor fell 8.8 minutes, from 14.3 minutes to 5.5 minutes. In addition, the mean wait to be called by the nurse fell from 20.1 minutes to 16.9 minutes, mean time spent with the provider declined by 4.6 minutes, and the mean postvisit processing time decreased by 4 minutes. The percentage of visits requiring hearing and vision screenings, immunizations, or laboratory work remained constant.

In a multivariate analysis, we regressed total visit time on the type of visit (routine or emergency); the type of practitioner (resident or attending pediatrician); the need for immunizations, laboratory studies, and hearing and vision screenings; and the year of the visit (Table 4). This analysis indicates that, holding other variables constant, the extra time added to a visit for a routine visit compared with an emergency visit was 5 minutes. Visits by residents were a little more than 11 minutes longer than those conducted by attending pediatricians. Immunizations, laboratory tests, and hearing and vision screenings added about 7, 18½, and 14 minutes to an average visit, respectively. Finally, holding all these elements constant, the decrease in visit time from 1999 to 2000 was a little more than 13½ minutes.

Table 4. 
Multivariate Analysis of Total Visit Time*
Multivariate Analysis of Total Visit Time*

In Figure 4 and Figure 5, 2 graphs depict changes in resident satisfaction with precepting and practice quality indicators for 1999 and 2000. Figure 4 shows that changes in the residents' perception of the availability of preceptors varied little during the 2 years of the study. The PL-1 trainees in 1999 were PL-2s in 2000, and PL-2s in 1999 were PL-3s the following year. Looked at in this context, interns in the first year felt a modest increase in the availability of preceptors in their second year, while second-year residents recorded a barely perceptible decline in this metric.

Figure 4. 
Resident perception of preceptor availability, 1999 and 2000. PL indicates postgraduate level.

Resident perception of preceptor availability, 1999 and 2000. PL indicates postgraduate level.

Figure 5. 
Selected quality indicators, 1999 and 2000. UTD indicates up-to-date; asterisk, P = .03.

Selected quality indicators, 1999 and 2000. UTD indicates up-to-date; asterisk, P = .03.

In Figure 5, selected quality indicators are shown for 1999 and 2000. These results were drawn from the continuous quality improvement audits conducted randomly each quarter throughout the year. Most indicators showed little variation from one year to the next. Immunization rates appear to have increased, although the magnitude of the increase did not reach statistical significance. The one improvement that did reach statistical significance was the number of charts that documented tobacco counseling at regular health care maintenance visits for parents of infants younger than 1 year.


We have shown that application of a time-flow study in a busy inner-city academic pediatric practice not only is feasible but can be a useful instrument to both analyze and exploit potential efficiency gains in practice patterns. In the particular application illustrated in this study, neither space nor the numbers of ancillary staff were found to be rate limiting in the processing of patients through the visit. Therefore, augmented space and staff were not necessary to achieve improvements in patient flow. Rather, modifications in the use of ancillary and attending pediatric staff proved to result in shorter patient waiting times.

The usefulness of the time-flow study resided in the clarity with which it identified portions of the visit where excessive waits appeared to occur. With the results of the time-flow study used as a guide, the medical, nursing, and administrative cadres at the practice were able to concentrate on problematic portions of the visit in the construction of potential solutions. Repeating the time-flow study then enabled the practice to test whether the implementation of these solutions resulted in improved waiting times. To avoid "squeezing the balloon" by improving one aspect of the practice at the expense of either the quality of training or patient care, the exercise also monitored changes in resident satisfaction and quality-of-care indicators.

It cannot be overemphasized that the ability to conduct both time-flow studies, to generate appropriate solutions, and to successfully implement the solutions depended critically on the collaboration of each segment of the health care team. The work to improve patient waiting times demonstrated to clerical, nursing, administrative, and medical personnel how complex an undertaking the ambulatory practice of pediatrics can be in a busy teaching facility, yet how effective a series of small changes can prove when derived collectively and linked together.

Several potential limitations of this analysis must be kept in mind when the results are interpreted. The data on time intervals were self-reported by each participating associate at the practice. Although clocks throughout the practice were synchronized, there was some room for inaccuracy in the estimation of time intervals. A previous study that compared direct observation of time intervals with self-reports among nurses found the former to be more accurate for certain types of nursing activity.9 Unlike the previous study, however, in the present analysis different individuals performed specific activities sequentially throughout the visit. With this design, each successive observer acted, in a sense, as a direct independent observer of the timing of the previous interval.

Moreover, a previous time-and-motion study conducted in a family practice residency teaching clinic observed similar waiting times.10 In that study, the authors documented a mean visit time of 80.5 minutes, of which 25 minutes was accounted for by registration and triage time, 19 minutes was spent waiting for a provider to begin, and a mean of 27 minutes was spent with a provider and preceptor. These times are quite similar to our own findings.

The possibility must still be considered that the subjects' awareness of the study may itself have affected the outcomes we observed. This eponymous Hawthorne effect11 might account for some of the distribution of times within each visit as a given participant attempted to minimize the time associated with his or her interaction. We believe, however, that since no incentives or disincentives were associated with the outcome of these observations, the motivation to falsify these time intervals was minimal. Furthermore, as other sequential time studies have pointed out,12 the Hawthorne effect would not account for changes over time in these intervals. A Hawthorne effect should have the same consequences for each of the 2 periods studied and, taking the difference between them, would net out this effect. For the Hawthorne effect to account for changes in the visit times across years, one would have to postulate an interaction between time and the Hawthorne effect—a contingency that we have no a priori reason to believe was obtained in this instance.

The room usage rates we calculated in the study are likely to be more accurate than the nursing usage rates, because a variety of nursing functions were not captured in this analysis. In addition to direct patient interactions as tabulated in this time-flow study, the nurses were also responsible for telephone triage functions, room stocking, filling out forms, sorting incoming laboratory studies, and other tasks that competed for their time with the patient-centered functions described in this study. Whether the reordering of these ancillary functions represents an additional source of potential gains in efficiency remains to be considered.

The relevant applicability of our findings lies not in the specific inefficiencies uncovered in the practice we evaluated, but rather in the technique used to discover those inefficiencies. If, as has been suggested elsewhere, ambulatory sites with teaching components have higher operating costs than nonteaching sites,13,14 then the motivation to uncover and exploit potential inefficiencies may be greater for ambulatory teaching programs than for comparable nonteaching entities. If the type of low-cost, easily administered time-flow study we present herein proves to be replicable in a variety of settings, then the use of such studies with the collaboration of clerical, nursing, and medical personnel will provide a useful instrument with which to improve the ability of ambulatory teaching sites to deliver more efficient care.


Accepted for publication July 11, 2002.

Corresponding author and reprints: Andrew D. Racine, MD, PhD, Division of General Pediatrics, Albert Einstein College of Medicine/The Children's Hospital at Montefiore, 1625 Eastchester Rd, Bronx, NY 10461 (e-mail:

What This Study Adds

Very little is known about the feasibility of applying time-flow techniques to the evaluation of patient waiting times in busy urban academic ambulatory pediatric practices. Whether such techniques can successfully identify potential time inefficiencies so that they can be reduced and then reevaluated without sacrificing the quality of teaching and clinical care has not previously been determined.

This study successfully demonstrates how the use of time-flow studies can identify areas of practice where patient waiting times can be improved and how the information disclosed by these studies can be exploited to reduce patient waits. It suggests a model of collaborative planning that involves all elements of the practice and shows that decreased waiting times can be achieved without sacrificing either the satisfaction of residents with their training or the ability of the practice to deliver high-quality care.

Schwartz  WBNewhouse  JPWilliams  AP Is the teaching hospital an endangered species?  N Engl J Med. 1985;313157- 162Google ScholarCrossref
Kassirer  JP Academic medical centers under siege.  N Engl J Med. 1994;3311370- 1371Google ScholarCrossref
Mechanic  RColeman  KDobson  A Teaching hospital costs: implications for academic missions in a competitive market.  JAMA. 1998;2801015- 1019Google ScholarCrossref
Reuter  JGaskin  D Academic health centers in competitive markets.  Health Aff (Millwood). 1997;16242- 252Google ScholarCrossref
Blumenthal  DWeissman  JSGriner  PF Academic health centers on the front lines: survival strategies in highly competitive markets.  Acad Med. 1999;741038- 1049Google ScholarCrossref
Iglehart  JK Rapid changes for academic medical centers, I.  N Engl J Med. 1994;3311391- 1395Google ScholarCrossref
Kassirer  JP Redesigning graduate medical education—location and content.  N Engl J Med. 1996;335507- 509Google ScholarCrossref
Task Force on the Future of Pediatric Education, The future of pediatric education, II: organizing pediatric education to meet the needs of infants, children, adolescents, and young adults in the 21st century: a collaborative project of the pediatric community.  Pediatrics. 2000;105 ((1, pt 2)) 157- 212Google Scholar
Burke  TAMcKee  JRWilson  HCDonahue  RMJBatenhorst  ASPathak  DS A comparison of time-and-motion and self-reporting methods of work measurement.  J Nurs Adm. 2000;30118- 125Google ScholarCrossref
Xakellis Jr  GCBennett  A Improving clinic efficiency of a family medicine teaching clinic.  Fam Med. 2001;33533- 538Google Scholar
Roethlisgerger  FJDickson  WJ Management and the Worker: An Account of a Research Program Conducted by Western Electric Company, Hawthorne Works, Chicago.  Cambridge, Mass Harvard University Press1939;
Kyriacou  DNRicketts  VDyne  PLMcCollough  MDTalan  DA A 5-year time study analysis of emergency department patient care efficiency.  Ann Emerg Med. 1999;34326- 335Google ScholarCrossref
Hogan  AJFranzini  LBoex  JR Estimating the costs of primary care training in ambulatory settings.  Health Econ. 2000;9715- 726Google ScholarCrossref
Linna  MHakkinen  ULinnakko  E An econometric study of costs of teaching and research in Finnish hospitals.  Health Econ. 1998;7291- 305Google ScholarCrossref