Observational study to determine predictors of rheumatology clinic visit provider contact time


  • The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs.



To address perceived inefficiencies in an academic rheumatology practice, a timing/work-flow evaluation was initiated to determine the factors that predict the provider contact time (PCT), i.e., the amount of time that attending physicians spend with patients during an outpatient encounter.


This prospective observational study was conducted at the University of Colorado Hospital Rheumatology Clinic for return patient visits in early 2008. Each patient encounter was subdivided into components, and the time for each component was recorded. Up to 20 return-visit encounters per provider were randomly selected for inclusion. Multivariate linear regression was used to predict the time, in minutes, that providers spent with patients, and logistic regression was used to determine the time intervals associated with the patient's perception that the visit ran on time.


Variables associated with increased PCT were whether a procedure was performed in the clinic (P = 0.037) and whether the visit occurred in the afternoon (P < 0.025). For every minute a provider was late in beginning to see a patient, the PCT decreased by 0.32 minutes (95% confidence interval [95% CI] 0.15, 0.49). Variables associated with the patient's perception that the visit ran on time included the check-in to vitals delay (odds ratio [OR] 0.95; 95% CI 0.92, 0.99) and the provider delay (OR 0.92; 95% CI 0.86, 0.99).


The patient's punctuality and the presence of a resident are not significantly associated with the time that a provider spends with a patient. However, the degree to which the provider runs late was associated with decreased PCT and diminishes the patient's perception that the visit is running on time.


The length of time that providers spend face-to-face with patients has been associated with patient satisfaction in outpatient visits (1, 2), and may, in fact, represent the most significant predictor of patient satisfaction (3). However, time pressures from various sources, including clinic inefficiencies and patient tardiness, may all influence the amount of time that a provider can spend with each patient. In addition, previous studies have implicated the presence of teaching residents with increased duration of the total clinical visit in both inpatient and outpatient settings (4, 5); however, it is unclear whether teaching residents affect specific components of the visit, such as the time the attending physician spends addressing patient issues.

Very few studies in the medical literature have examined total clinic visit durations, much less the time required for components of visits. A small number of studies have examined the amount of time spent in each part of a patient visit for various medical specialties (6–8). A review of the literature, however, does not reveal any published data reporting clinic visit times for rheumatology clinics, nor have prior studies evaluated predictors of the amount of time that attending physicians spend with patients during clinic visits (provider contact time [PCT]).

Data were collected in order to satisfy 3 objectives: 1) describe the duration of each component of rheumatology clinic encounters as well as the duration for the entire encounter, 2) determine which clinic characteristics (such as the presence of resident physicians) predict the PCT, and 3) determine which time intervals were associated with the patient's perception that the visit ran on time.

Materials and Methods

Study design, setting, and ethics.

A prospective observational study design was used to identify predictors of PCT and predictors of patient's perception that the visit ran on time. The study was conducted at the University of Colorado Hospital Rheumatology Clinic and was based on clinical encounters from January 15, 2008 to March 4, 2008. At the time of the study, 10 faculty members staffed this academic clinic.


All followup encounters with adult patients presenting to the clinic were tracked in time and were eligible to be included in the analysis. All providers in the rheumatology clinic participated. From among the encounters occurring during the period of observation, a random sample of ≤20 encounters per provider was selected for analysis. The limited number of initial patient consultations and variations in the structure of these new patient visits precluded analysis of these visit types; these encounters were excluded. All data were collected as part of a performance improvement project initiated by the administrative staff in the rheumatology division of the University of Colorado Hospital. Accordingly, a patient waiver of consent and waiver of authorization were obtained.

Data sources/measurement.

The scheduled visit time for each patient was abstracted from administrative records. Seven additional times (check-in, vitals start, vitals end, provider start, provider end, phlebotomy end, and check-out) were recorded on preprinted timing cards through the course of a clinical visit (Figure 1). An automated time stamp that could not be adjusted by the staff stamped the check-in and check-out times on the cards. Interim visit times were recorded in ink by the staff responsible for each component and confirmed by the staff responsible for the subsequent component. Medical records were reviewed to confirm whether patients underwent phlebotomy or procedures.

Figure 1.

Times recorded (in minutes) during patient encounters and the intervals calculated from those recorded times. * = Check-in time could occur either before (patient arrives early) or after (patient arrives late) the scheduled visit time; ** = outcome variable.


Time-related variables.

The number of intervals were calculated from the scheduled visit time and 7 subsequent times (Figure 1). Administrative staff selected these intervals by determining the points in time at which one health care team member completed his or her responsibilities, thereby allowing a subsequent health care team member to proceed. These intervals included: 1) arrival interval (patient check-in time to scheduled visit time; this may be positive [late arrival] or negative [early arrival]), 2) check-in to vitals delay (wait times between check-in time and when vital signs are obtained by a medical assistant), 3) vitals duration, 4) provider delay (wait time between completion of vital signs and provider start time), 5) PCT, 6) phlebotomy interval (time from the completion of the provider visit until the completion of phlebotomy by the medical assistant), 7) total visit duration (time from the patient's scheduled visit time to check-out), and 8) patient time in clinic (time from check-in to check-out). The PCT represents the time the provider attends to the patient. For visits in which a resident participated in the care of the patient, the PCT did not include the time the resident spent gathering information, interviewing the patient, and performing the physical examination, but started as soon as the resident began presenting to the attending physician. The PCT includes the attending physician's precepting time and face-to-face time with the patient, and ends when the attending physician sends the patient for laboratory testing or for discharge from the clinic. The phlebotomy to check-out interval was not included, since this interval was very short. All intervals were calculated in minutes.

Additional variables.

We recorded whether the patient was seen by a resident (rheumatology fellows' clinics were not studied), whether the visit was scheduled for an afternoon appointment, whether a procedure (arthrocentesis or glucocorticoid injection) was performed, and whether phlebotomy was obtained at the visit. Patients whose laboratory tests were ordered for outside of our university were not included in this variable. All of these additional characteristics were modeled as binary variables. We incorporated a comorbidity index based on the patient's administrative data (9). To mitigate potential bias resulting from changes in efficiency with time, the number of days since the initiation of the study observation period was included as a control variable.

Outcome measures.

The PCT served as the outcome measure. For the analyses to determine which time intervals were associated with the patient's perception that the visit ran on time, we asked the patient to complete a simple written (yes/no) question after they had checked out: “Did your visit run on time today?”

Statistical analysis.

We performed multivariate linear regression to determine whether the amount of time for each component interval and the characteristics of each encounter predicted the PCT. Unconditional logistic regression was employed to determine which time intervals were associated with the patient's perception that the visit ran on time. The models were constructed by initially employing univariate regression to identify those predictors in which P was < 0.10. Variables found to be significant in univariate analysis were incorporated into multivariate regression, and for this analysis, P values less than 0.05 (2-sided) with 95% confidence intervals (95% CIs) were considered significant. The multivariate regressions accounted for the lack of independence among observations for each provider. All analyses were performed using Stata statistical software, version 10.2 (StataCorp).


Descriptive data.

Characteristics for the established patient encounters appear in Table 1. A total of 191 established patient encounters were available for analysis. Patients spent, on average, 24 minutes in PCT. They averaged >75 minutes in the clinic and arrived, on average, 15 minutes prior to their appointment time. Patients waited an average of ∼11 minutes due to provider delay.

Table 1. Descriptive characteristics*
  • *

    Values are the mean ± SD unless indicated otherwise.

Patient's time in clinic (check-in until check-out)19175.40 ± 32.06
Total visit duration (scheduled visit time until check-out)19160.61 ± 31.60
Check-in to vitals delay18015.23 ± 16.31
Provider delay11410.87 ± 13.99
Arrival interval (check-in time, relative to scheduled visit time)191−14.79 ± 21.81
Vitals duration1745.59 ± 2.21
Provider contact time11724.07 ± 11.36
Phlebotomy interval1445.63 ± 6.71
Procedure performed during visit, %1915
Phlebotomy performed during visit, %19180

Outcome data.

Results from the univariate and multivariate regression appear in Table 2. The statistically significant variables associated with PCT in the multivariate analysis included: provider delay, whether a procedure was performed, and whether the visit was scheduled for an afternoon appointment. PCT was decreased with increased provider delay (P < 0.001). For every minute a provider was late in beginning to see a patient, the PCT was decreased by 0.32 minutes (95% CI 0.15, 0.49). The patient's punctuality and the presence or absence of a resident physician was not associated with PCT. When the provider performed a procedure, the PCT was 13.15 minutes longer (95% CI 0.78, 25.52). In afternoon visits, the PCT was predicted to be 5.07 minutes longer than those in the morning, controlling for other differences (95% CI 0.64, 9.50). Only 2 variables were associated with the patient's perception that the visit ran on time: the check-in to vitals delay (OR 0.95, 95% CI 0.92, 0.99; P = 0.008) and the provider delay (OR 0.92, 95% CI 0.86, 0.99; P = 0.019).

Table 2. Predictors of provider contact time, as determined by multivariate linear regression*
VariableUnivariate analysisMultivariate analysis
CoefficientP95% CICoefficientP95% CI
  • *

    95% CI = 95% confidence interval.

  • Controlled for the fact that observations contributed by the same provider are not independent.

  • Significant value.

  • §

    Modeled as binary variables, where yes = 1 and no = 0.

Days since initiation of study−0.0010.991−0.190, 0.187   
Arrival interval0.0360.534−0.078, 0.151   
Check-in to vital delay0.0270.750−0.142, 0.196   
Vitals duration0.8520.172−0.376, 2.080   
Provider delay−0.2920.004−0.490, −0.093−0.319< 0.001−0.491, −0.148
Phlebotomy interval0.3830.0430.012, 0.7550.0350.818−0.263, 0.332
Resident physician during visit§1.9240.657−6.638, 10.487   
Afternoon appointment§6.2980.0171.134, 11.4625.0700.0250.644, 9.496
Phlebotomy obtained§−1.6270.650−8.699, 5.445   
Procedure performed§19.4300.0036.705, 32.15413.1470.0370.778, 25.517
Comorbidity score−3.3830.020−6.225, −0.5420.2690.807−1.913, 2.452


If the deficit in the rheumatology workforce continues to widen as predicted (10), the need to understand the causes of clinical inefficiency will become more acute. As an initial step toward this objective, our report, to our knowledge, provides the first descriptive information for the times that comprise a rheumatology visit for established patients.

The longer the patient had to wait between the completion of their vital sign assessment and the provider beginning to attend to the patient (i.e., the provider delay), the shorter the PCT. This may indicate that the provider ran late for these encounters and was trying to “make up time” by decreasing the time they spent with the patient. Since the delay due to providers averaged 10.81 minutes, and the providers spent 0.32 minutes less attending to the patient for each minute of this delay, PCT was reduced an average of 3.5 minutes for visits with a provider delay. Patients were more likely to report that the visit did not run on time in visits with a provider delay. Although we did not obtain patient satisfaction scores, the combination of a longer wait time and shorter time with the physician has the potential to negatively influence patient satisfaction. Anderson et al argue that because the time spent with the physician may be a much stronger predictor of patient satisfaction than waiting time, shortening patient waiting times at the expense of time spent between the patient and the physician would be counter-productive (3). Since patient experience is emerging as a more important outcome in quality-based reporting and reimbursement, providers should recognize the consequences of keeping patients waiting.

This study's findings resulted in a number of practice redesign processes. Medical assistants (MAs) now obtain vital signs soon after patients are checked in, rather than waiting for an available examination room. Because afternoon visits were associated with 5 minutes longer PCT than morning visits, the scheduled work hours of the MAs are now staggered to better reflect patient volumes. Providers spent >13 minutes more time caring for the patient when an arthrocentesis or glucocorticoid injection was performed. To streamline procedures, providers are now assigned a specific MA for each clinic (rather than working with the next available MA). This last change enhances communication between the staff, so that providers are immediately aware when a patient is roomed and also allows a more rapid set-up for procedures.

Of particular interest, however, is our inability to demonstrate a relationship between the presence of a trainee and the PCT. Trainees may not affect the time devoted to each patient by the attending physician because they have the potential to collect, synthesize, and recapitulate basic clinical information for the attending physician, although this effect may vary by year of resident training and does not necessarily reflect physician productivity (11). The efficiency associated with residents may be counterbalanced by the need for the attending physician to review clinical information with the resident and to teach. The fact that trainees may be time-neutral with respect to the attending physician's clinical efficiency has implications for calculating the expense of operating a training program. Typically, such studies account for teaching time without considering the benefit to the attending physician's clinical efficiency (12). Similarly, patient tardiness was not associated with PCT. It would appear that in our study providers did not “penalize” patients for being late by abbreviating their time with the patient.

This study has several limitations. The sampled encounters were restricted to a 7-week period and may not reflect relationships throughout the year. Our method of recording interim times throughout each encounter would not identify inaccurate times, except for the check-in and check-out times, which were recorded by an automated time stamp. Also, providers and staff were not blinded to these values during each encounter. Although we included patient comorbidity as a variable in the analysis, additional factors related to the medical content of the visit (e.g., the severity of specific illnesses, relationship-building, and patient education) were not measured and may have influenced the results. Last, although our study included all patients attending an academic rheumatology clinic, its relevance to community practice remains uncertain.

Very few prior studies address the issue of timing and workflow through the clinic; to our knowledge, none provide data derived from rheumatology clinical encounters. While it is not possible to determine whether the particular findings of this study may be extrapolated to all rheumatology practices, the methodology described may serve as the initial template for conducting future timing studies in different settings.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Dr. Caplan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Caplan.

Acquisition of data. Larson, Caplan.

Analysis and interpretation of data. Davis, Caplan.


The authors wish to thank Judith Potter for initiating this quality improvement project.