Fast and Frictionless: A Novel Approach to Radiology Appointment Scheduling Using a Mobile App and Recommendation Engine

Abstract

Many outpatient radiology orders are never scheduled, which can result in adverse outcomes. Digital appointment self-scheduling provides convenience, but utilization has been low. The purpose of this study was to develop a “frictionless” scheduling tool and evaluate the impact on utilization. The existing institutional radiology scheduling app was configured to accommodate a frictionless workflow. A recommendation engine used patient residence, past and future appointment data to generate three optimal appointment suggestions. For eligible frictionless orders, recommendations were sent in a text message. Other orders received either a text message for the non-frictionless app scheduling approach or a call-to-schedule text. Scheduling rates by type of text message and scheduling workflow were analyzed. Baseline data for a 3-month period prior to the launch of frictionless scheduling showed that 17% of orders that received an order notification text were scheduled using the app. In an 11-month period after the launch of frictionless scheduling, the rate of app scheduling was greater for orders that received a text message with recommendations (frictionless approach) versus app schedulable orders that received a text without recommendations (29% vs. 14%, p < 0.01). Thirty-nine percent of the orders that received a frictionless text and scheduled using the app used a recommendation. The most common recommendation rules chosen for scheduling included location preference of prior appointments (52%). Among appointments that were scheduled using a day or time preference, 64% were based on a rule using the time of the day. This study showed that frictionless scheduling was associated with an increased rate of app scheduling.

Introduction

Outpatient radiology examinations provide important diagnostic information for the detection and monitoring of disease processes. After a healthcare provider orders an imaging examination, the onus of contacting an imaging center to schedule is usually upon the patient. Potential barriers to scheduling over the phone include identification of an imaging center that performs the examination, locating the contact information, reaching a scheduler during business hours, complicated phone trees, transfers, or hold times. This may result in frustration and abandonment of the scheduling process. Lacson et al. found that 7% of radiology orders were never scheduled [1]. Failure to complete imaging can result in delayed diagnosis and adverse outcomes.

Digital self-scheduling of appointments can help overcome barriers to traditional phone scheduling. A prior report found that in 2018, 16.9% of patients used the internet to schedule a medical appointment [2]. Online appointment systems can also reduce no-show rates, waiting times, costs, and staff labor while improving satisfaction, efficiency, popularity, and revenue [3, 4]. Potential barriers to use of digital scheduling include access to computers or internet, multiple input fields, screen prompts, choices, or required clicks, as each additional step presents a potential point of abandonment. In other industries, excess choice can make it more difficult for customers to choose a product, and recommender systems can help [5]. Prior research suggested that typical Netflix members lose interest after 60–90 seconds of choosing and reviewing 10–20 titles on 1–2 screens [6].

Prior internal experience with an app-based radiology scheduling tool revealed that 17% of orders that received a text notification to schedule on the app were self-scheduled. The purpose of this initiative was to develop a seamless scheduling tool that increases the rate of app-based scheduling by reducing potential points of friction.

Methods

The study qualified as a quality improvement project and did not require institutional review board review.

Clinical Setting

Our radiology department performs approximately 2 million exams per year and provides services at 6 hospitals and over 50 outpatient sites.

Radiology Scheduling App

In 2018, a custom-developed radiology portal was added to the institutional patient mobile health app, built upon MyChart software tools and interfaces made available by Epic (Epic, Verona, WI). The mobile app portal included a scheduling tool that allowed direct app scheduling of most outpatient diagnostic radiology exams that were ordered digitally within the electronic health record (EHR) system, including X-ray, mammography, ultrasound, CT, MR, and FDG PET-CT. Nuclear medicine exams, PET/MR, procedures, and certain other exams that required special consideration (i.e., those that required scheduling at a particular facility or time based on imaging protocol) were excluded. Order notification text messages included an “app schedulable” message with a link to the app for exams eligible for app scheduling or a “call-to-schedule” message that included the phone number for a scheduling center for those that could not be scheduled on the app or for patients without active portal accounts.

Frictionless Scheduling Initiative

To enable frictionless scheduling, several software developments and process changes were necessary.

Expected Date

Providers order some exams with the intention that they be scheduled on a certain future date. For example, a chest CT that reveals pulmonary nodules may require a follow-up chest CT in 3 months to assess for stability. The ordering provider may order the CT at the time of the initial result and inform the patient to schedule it for a date 3 months in the future. The frictionless scheduling system requires knowledge of such an intended timeframe in order to recommend a clinically appropriate appointment date. This was achieved through utilization of an existing field in the EHR called “expected date,” which was set to be required for all future outpatient diagnostic imaging orders except for radiographs.

Eligible Exams

Outpatient exams ordered digitally in the electronic health record (EHR) were eligible for frictionless scheduling. Most mammography, ultrasound, X-ray, CT, MR, and FDG PET-CT examinations were included. Only examinations that had a value in the “expected date” field were included, and only exams with expected dates less than 90 days in the future were eligible due to potential performance issues when including more extended periods. Patients were required to have an active patient portal account and the mobile app to schedule using the frictionless approach. Only single radiology orders were eligible for frictionless.

Recommendation Algorithm

A recommendation algorithm was developed to predict optimal appointment location, date, and time. The engine utilized data including patient residence zip code, location, time and day of week of prior imaging appointments and future radiology, or other clinic appointments. All appointments up to 24 months in the past and 6 months in the future were evaluated. Patient appointment preferences were determined by identifying strong dominant, weak dominant, or non-dominant trends in location (city/town and zip code), appointment day of the week, and time of day. Preferences were determined based on appointment types in the following priority order:

• Most recent past radiology appointment

• All other past radiology appointments

• All other past non-radiology appointments

• Next future radiology appointment

• All other future radiology appointments

• All other future non-radiology appointments

Example 1

A patient has 5 past radiology appointments—1 is in the morning, 1 is in the afternoon, and 3 are in the evening. The most recent one is in the evening.

The evening is a weak preference (60%), but it also has the highest weight as the most recent appointment, so we determine the patient has strong evening preference.

Example 2

A patient has 9 past appointments—3 non-radiology are on a Tuesday, 3 non-radiology are on a Friday, and 3 radiology are on a Wednesday.

We determine the patient has a weak preference for Wednesday.

Next, the engine searches for recommendations based on the following criteria and trends:

• Recommendations based on future appointments that fall within the expected date search window

• Recommendations based on past appointment preferences

• Recommendations based on home address

• Recommendations based on ordering department location

Up to two recommendations were based on strong dominant trends. Only one recommendation was based on a weak dominant trend. If day and time are dominant, one recommendation will be based on day preference and another will be based on time.

The recommendations are put through various scheduling filters and restrictions. Additional rules were built into the algorithm to account for CT, MRI, and FDG PET-CT, which may require prior authorization based on insurance providers. If prior authorization was required, a buffer of 5 business days was added for MRI and CT. If no prior authorization was required, a 3–4-h buffer was added before the first shown available appointment time to allow enough time for patient preparation, including fasting, if indicated. When prior authorization is approved, patients are automatically sent a message through their portal. If prior authorization is denied, staff reach out to the patient to adjust scheduling.

The engine also attempted to provide a unique imaging site for each recommendation, but they could all be the same after applying all filters and restriction rules. The order of the recommendations was randomly changed so that each recommendation was based on a different rule. The engine did not offer any appointments within 2 hours of any future appointment to avoid overlap.

Patient Notification

Patients who had communication preference settings that permitted SMS text messaging were sent order notification messages. Text messages were sent on weekdays in one of three batches, including 10:30 am, 1:30 pm, or 4:30 pm based on the ordering time. Before the implementation of frictionless, there were two types of text messages: an “app schedulable” message for exams eligible for app scheduling that provided a link to the app and a “call-to-schedule” message that provided a phone number to a scheduling center for exams not eligible for app scheduling. For the frictionless scheduling initiative, a third type of text message was developed that included details of the three recommended appointment types with a link to log into the app. Samples of the text message types are shown on Fig. 1.

Fig. 1.

Sample order notification text messages

App Design Updates

For radiology orders that are not eligible for frictionless scheduling, a “deep-link” in the text message directs patients to the radiology portal of the app after login, where their radiology orders available for digital scheduling are displayed. The patient can begin the scheduling process by tapping a button under the exam name. For the frictionless scheduling approach, the “deep-link” in the text message adds the additional functionality of pre-selecting the individual order. A new interface was created to accommodate the frictionless workflow (Fig. 2), which displays the examination name, ordering provider, order date, and the option to swipe through the three recommended appointments or to access the full appointment calendar.

Fig. 2.

Sample view of the frictionless scheduling screen on the app

Once an appointment is selected, the patient must answer “scheduling questions” that evaluate for any factors that would require a call with a scheduler to ensure that the exam is scheduled safely and appropriately (for example, contrast allergies). If all the answers permit app scheduling, the patient proceeds to the confirmation screen to complete scheduling, upon which the appointment details are shown, including maps, preparation instructions, and the option to complete questionnaires.

Data Analysis

A dashboard was developed to monitor app scheduling rates. Baseline scheduling data was analyzed for a 3-month period (January 2021–March 2021) before the launch of frictionless scheduling. After the launch of frictionless scheduling, data was analyzed for orders placed May 2021 to March 2022. Descriptive statistics were computed using Microsoft Excel. Statistical analysis was performed using the N − 1 chi-squared test for comparison of proportions using MedCalc (MedCalc, Ostend, Belgium).

Results

Baseline data for a 3-month period before the launch of frictionless scheduling showed that 17% (11,140/67,206) of orders that received an app schedule text message were scheduled using the app.

During the 11-month study period with frictionless scheduling, 20% (303,460/1,521,158) of schedulable orders received an order notification text message. Among these, 38% (114,206/303,460) were frictionless texts with recommendations, 40% (121,200/303,460) were app schedulable texts without recommendations, and 22% (68,054/303,460) were call-to-schedule texts without recommendations.

The rates of scheduling method by type of text message are provided in Table 1. The rate of scheduling on the app was greater for orders that received frictionless texts (29%) than those that received app schedulable texts (14%) p < 0.01. Of the orders that received a frictionless text and scheduled using the app, 39% (13,228/33,632) used a recommendation.

Table 1.

Scheduling rates by type of text message

Type of text message	Number	Scheduled on the app	Scheduled without using the app
Frictionless (with recommendation)	114,206	33,632 (29%)	37,621 (33%)
App schedulable (no recommendation)	121,200	17,182 (14%)	51,451 (42%)
Call-to-schedule (no recommendation)	68,054	1411 (2%)	42,277 (62%)

Table 2 presents the rates of app scheduling by message type and exam modality. The rate of app scheduling for all modalities was highest for orders that received a frictionless text message, ranging from 9% for PET/CT to 37% for MRI.

Table 2.

Rates of app scheduling by type of text message and modality

Type of text message	X-ray	US	Mammography	CT	MR	PET/CT
Frictionless (with recommendation)	25% (2,112/8,548)	37% (9,304/25,009)	24% (6,163/26,216)	28% (9,660/34,090)	32% (6,316/19,501)	9% (77/842)
App schedulable (no recommendation)	11% (2,209/20,910)	22% (1,841/8,228)	14% (9,171/63,945)	12% (1,894/15,309)	16% (2,058/12,582)	4% (9/226)
Call-to-schedule (no recommendation)	1% (36/4,264)	2% (136/8,232)	3% (318/10,785)	2% (68/4,383)	2% (836/39,431)	2% (17/959)

Recommendation types used for scheduling were analyzed for a 7-month period from May–November 2021. During this period, 15,245 appointments were scheduled using a recommendation (Table 3). Fifty-two percent (7,926/15,245) were scheduled using a rule that included a location preference of prior appointments (city, zip code, or department), 28.9% (4,401/15,245) were based on an imaging center within 5 miles of the patient’s home address, and 13.4% (2,039/15,245) were based on the ordering department location. A total of 4,931 of the recommendations included a day of week and/or time of day preference. Among these, 63.7% (3,142/4,931) were scheduled using time of day, 17.2% (850/4,931) using day of week, and 19.0% (939/4,931) using day of week and time of day.

Table 3.

Appointments scheduled by recommendation type

Rule	Number of scheduled appointments	Percent of total scheduled appointments
Prior appointment city		35.1% (5,358/15,245)
Without day/time preference	1,992
+ Day of week	622
+ Time of day	2,118
+ Day of week and time of day	626
Prior appointment zip code		15.0% (2,288/15,245)
Without day/time preference	1,062
+ Day of week	165
+ Time of day	836
+ Day of week and time of day	225
Prior appointment department		1.8% (280/15,245)
Without day/time preference	46
+ Day of week	40
+ Time of day	114
+ Day of week and time of day	80
Earliest available appointment	770	5.1% (770/15,245)
Patient home address	4,401	28.9% (4,401/15,245)
Ordering department	2,039	13.4% (2,039/15,245)
Day/time only	105	0.7% (105/15,245)
Future appointment	4	0.0% (4/15,245)
Total	15,245

Discussion

Frictionless scheduling increased the rate of app scheduling from 17% at baseline to 29%, indicating that the frictionless scheduling approach is an effective tool to help promote app scheduling. The frictionless approach eliminated the steps required to find an imaging location, day, and time, which are areas where patients may abandon the scheduling process. For example, the process of finding a location begins with entering a zip code, but the patient may not know the zip code of a particular neighborhood or town where they want to get their imaging. Next, the calendar system requires advancing by individual weeks, and patients may abandon the process if the imaging date is weeks or months into the future. While 39% of the orders that received frictionless texts were scheduled digitally using one of the offered recommendations, the remainder were scheduled by manually choosing a location and date. Although 61% of the patients did not use one of the recommended appointments, the presentation of the recommendations may have helped inform them of available imaging centers and hours. The frictionless approach also may have demonstrated the ease of scheduling.

The recommendation algorithm leveraged several data points to help offer optimal recommendations. The Netflix recommender system aimed to ensure that the first two screens display an optimal item for each member and influenced 80% of streaming hours [6]. Evaluation of prior patient preferences is useful to help generate recommendations. Patients may prefer to schedule their appointments closer to their workplace than their home, in which case data on prior appointments would be more helpful than residence zip code. Additionally, there are several imaging locations within the same zip code within close proximity to each other, but patients may prefer a particular site for a variety of reasons. Patients may need to get imaging tests completed prior to their next clinic appointment, in which case future appointment location and date information is useful.

Digital radiology appointment scheduling offers a convenient approach to help overcome traditional barriers to scheduling. Satisfaction surveys showed that online appointment scheduling is an important feature and that most patients would use such a service again [4]. There are several advantages, including the ability to schedule outside of normal business hours. A study of digital screening mammography self-scheduling found that 24.4% of self-scheduling activity occurred outside of normal business hours, including 15.9% on weekdays and 8.5% on weekends [7].

Additional advantages include a method to transmit helpful information, such directions with maps and preparation instructions. Downstream benefits include the opportunity for patients to complete registration forms, which improve throughput when they arrive for their appointments. Internal radiology scheduling call center data revealed that the average call duration for scheduling a radiology appointment over the phone was 3.5 minutes. Approximately 10,000 appointments are scheduled on the app monthly, corresponding to 462 appointments per business day, which could save 27 scheduler hours per day. App scheduling can help reduce scheduler staffing requirements, decrease the need to hire as volumes increase, and provide opportunities for schedulers to focus on other initiatives.

Despite the convenience of digital scheduling, the overall utilization remained low. Similarly, North et al. found that 15.3% of patients performed digital self-scheduling or canceling activity for screening mammogram orders [7]. Barriers to use of digital scheduling include lack of awareness, lack of trust in the computer system, low computer literacy, lack of access to the internet, or preference to discuss options for complex situations over the phone [8]. For radiology exams, it is possible that patients preferred to speak with a scheduler to obtain exam information, insurance authorization requirements, review preparation instructions, or ensure correct scheduling. North et al. found that 75.5% of self-scheduling activity was on the web while 24.5% was on a mobile app [7]. Radiology appointment self-scheduling at our center is currently only available on the mobile app. It is possible that offering a web option would result in a greater rate of digital scheduling.

This study has several limitations. There were no direct patient interviews or surveys to better understand why patients do not prefer to use app scheduling or reasons for abandoning the scheduling process. We did not evaluate how many of the app scheduled appointments resulted in completed exams, cancelations, no-shows, or rescheduling activity by a scheduler. Frictionless scheduling was only available for expected date less than 90 days and the relationship between timing of the expected date and app scheduling was not evaluated. The “deep-link” provided with the frictionless scheduling text messages pre-selected the order to be scheduled, while the “deep-link” provided with the non-frictionless texts required a tap to select the order. It is possible that this extra tap could be a source of abandonment of the scheduling process.

In conclusion, frictionless radiology scheduling has the potential to increase the rate of digitally scheduled appointments, decrease scheduler staffing requirements, and improve patient convenience and satisfaction. Future investigations could focus on optimization of the recommendation engine or use of machine learning.

Author Contribution

All authors contributed to the study conception and design. The first draft of the manuscript was written by Ankur Doshi and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Declarations

Ethics Approval

The study qualified as a quality improvement project and did not require institutional review board review.

Conflict of Interest

The authors declare no competing interests.