Using Self-Scheduling to Improve Screening Mammography Completion Rates

Abstract

Purpose:

Self-scheduling has the potential to enhance convenience and patient engagement. We compared outpatient screening mammography completion rates before and after implementing an online self-scheduling system between patients who use self-scheduling versus traditional scheduling.

Methods:

In February 2021, a self-scheduling process was implemented at an institutional level through the Epic MyChart online portal, allowing patients to self-schedule screening mammography. This retrospective cohort study included women aged 18 and over who scheduled outpatient screening mammography in a tertiary health care facility from October 1, 2017, to June 30, 2023, had at least one encounter during the pre-implementation phase and one encounter during the postimplementation period, and only used one scheduling method (self-scheduling or traditional scheduling) in the postimplementation period. Difference-in-difference analyses were conducted to compare screening mammography completion rates between patients who used traditional versus self-scheduling in the postimplementation period.

Results:

In all, 29,893 screening mammography were scheduled by 7,203 patients (mean age: 58.1 years; 70.0% White, 18.2% Asian, 1.8% Black, and 19.5% Hispanic). The overall mammography completion rate in pre-implementation period was 78.9% and increased to 79.8% in the postimplementation period. Using difference-in-difference estimator, the completion rates in the self-scheduling cohort was 8.4 percentage point (95% confidence interval, 5.2–11.6) higher than traditional scheduling. The change in screening mammography completion rate from the postimplementation to pre-implementation period was +8.5 percentage point (88.1% postimplementation versus 79.6% pre-implementation) for the self-scheduling cohort and +0.1 percentage point (80.8% postimplementation versus 80.7% pre-implementation) for the traditional scheduling cohort.

Conclusion:

Self-scheduling was linked to increased screening mammography completion rates postimplementation when compared to traditional scheduling.

Graphical Abstract

INTRODUCTION

Early detection of breast cancer significantly reduces morbidity and mortality [1–3]. Screening mammography, along with other imaging and clinical assessments, is crucial for detecting early breast cancer. However, its utilization is influenced by sociodemographic factors such as racial or ethnic background, insurance, and neighborhood socio-economic status [4,5].

To address these disparities, targeted interventions are needed. The increasing availability of technology to automate screening orders and facilitate self-scheduling with the aid of electronic medical records (EMRs) offers a cost-effective solution. In 2022, 60% of US adults accessed online medical records [6], making EMRs a valuable tool for distributing health information and allowing self-scheduling without need for additional staffing [7]. Self-scheduling has been shown to enhance patient satisfaction, accountability, and time savings [8]. By empowering patients through self-scheduling, radiology services can help promote patient autonomy, improve access to care, and foster a collaborative approach.

Previous studies on self-scheduling for screening mammography [9–12], diagnostic mammography [13], and diagnostic imaging [14] show varying utilization rates, from 0.7% in 2019 [10] and 36.9% in 2019 [9]. Certain groups, such as younger individuals, those of non-Black racial or ethnic background, and those whose primary language was English [9,10,12] were more likely to use self-scheduling. Research on association between self-scheduling and examination completion has yielded mixed results. Wadell et al found increased screening mammography completion after self-scheduling implementation [15], and Khiati et al reported fewer cancellations or no-shows among patients who self-scheduled their diagnostic mammography [13]. In contrast, North et al and Sadeghi et al found higher missed appointment rates among patients who self-scheduled screening mammography [11,12].

Although self-scheduling has the potential to enhance convenience and patient engagement, limited studies are available to assess its effectiveness in improving screening mammography completion rates. To address this, we compared screening mammography completion rates before and after implementing self-scheduling in our health system between patients who used self-scheduling versus traditional scheduling.

METHODS

The institutional review board approved this retrospective study, and consent and a HIPAA authorization waiver were obtained.

Study Population

In February 2021, a self-scheduling process was implemented at an institutional level through the Epic MyChart online portal, allowing patients to self-schedule screening mammography. Patients still had the option of using traditional scheduling by calling imaging center. Eligible patients were 18 years or older who were female at birth and had at least two scheduled outpatient screening mammography encounters at imaging centers affiliated with our institution, which is a tertiary academic center in Southern California. At least one of these scheduled examinations should have been during the self-scheduling pre-implementation period, which spanned from October 1, 2017, to January 31, 2021, and the second one should have been during the postimplementation period, which spanned from February 1, 2021, to June 30, 2023. We further limited the eligibility to patients who only used one type of scheduling method (ie, self-scheduling through online portal or traditional scheduling by calling the imaging center) in the postimplementation period. A list of eligible patients and encounters was obtained by querying institutional radiology data warehouse.

Data Measures

For each eligible encounter, the scheduled date, method of scheduling during the postimplementation period (self-scheduling or traditional scheduling), and appointment status (completed, cancelled, no-show) were extracted from EMR. The following sociodemographic variables were also extracted from EMR: sex at birth, age, primary language, race, ethnicity, marital status, health insurance provider, and address with zip code. We calculated the area deprivation index (ADI) national scores by mapping patients’ address using the University of Wisconsin Neighborhood Atlas. The ADI is an index of socioeconomic status ranging from 0 to 100 with higher scores equating to more disadvantaged groups [16].

The outcomes of interest were screening mammography completion (binary outcome) before and after implementing online self-scheduling.

Statistical Analysis

Analyses were completed in 2024. All patients were classified into self-scheduling versus traditional scheduling cohorts based on their use of self-scheduling in the postimplementation period. Descriptive statistics and summaries are reported for continuous and categorical demographic variables. Continuous variables are expressed as means ± SDs and categorical variables as frequencies and percentages.

Unadjusted analyses examined screening mammography completion rates by cohort. Changes in screening mammography completion rates from the pre-implementation period to the postimplementation period by scheduling pathway status were examined using a difference-in-difference design. An event study cohort was performed for each cohort to examine if the trends in screening mammography completion rates were similar in the pre-implementation period. Adjusted analyses fit generalized estimating equations, and adjusted for age at examination, race, ethnicity, implementation period, insurance type, and ADI. Language was not included in the model due to collinearity with race. The difference-indifference estimate was obtained using an interaction term for time (pre-implementation versus postimplementation period) × scheduling method. The delta method was used to convert the difference-in-difference estimate and confidence interval from the log odds scale to a percentage difference. Statistical analysis was performed with Stata/MP 18.0 software (StataCorp LLC, College Station, Texas) and SAS v9.4 (SAS Institute Inc., Cary, North Carolina). Statistical significance was set at P < .05.

RESULTS

During the study period, a total of 29,893 screening mammograms were scheduled by 7,203 patients who had at least one screening mammography encounter in both pre- and postimplementation periods and used only one of the scheduling methods during the postimplementation period. All patients used traditional scheduling during the pre-implementation period. Overall, 6,575 patients (91.3%) used only traditional scheduling in the postimplementation period, and 628 patients (8.7%) used only self-scheduling in the postimplementation period.

The mean patients’ age was 58.1 (SD 11.2). The racial background of patients was 18.2% (n = 1,310) Asian, 1.8% (n = 132) Black, 9.1% (n = 656) other races, 70.0% (n = 5,041) White, and 1.0% (n = 74) with unknown race. Further, 19.5% (n = 1,405) of included patients were of Hispanic or Latino ethnicity. Most patients had private insurance (54.0%; n = 3,892) and spoke English (85.4%; n = 6,152). Baseline characteristics are shown in Table 1.

Table 1.

Baseline characteristics of study population

Characteristics	Traditional Scheduling Cohort* (n = 6,575)	Self-Scheduling Cohort† (n = 628)	Total (n = 7,203)
Age (y), mean (SD)	58.3 (11.20)	55.8 (10.39)	58.1 (11.2)
Race, n (%)
Asian	1,147 (17.4)	163 (26.0)	1,310 (18.2)
Black	112 (1.7)	20 (3.2)	132 (1.8)
Other	598 (9.1)	58 (9.2)	656 (9.1)
Unknown	66 (1.0)	8 (1.3)	74 (1.0)
White	4,662 (70.9)	379 (60.4)	5,041 (70.0)
Ethnicity, n (%)
Hispanic	1,312 (20.0)	93 (14.8)	1,405 (19.5)
Non-Hispanic	5,197 (79.0)	528 (84.1)	5,725 (79.5)
Unknown	66 (1.0)	7 (1.1)	73 (1.0)
Language, n (%)
English	5,563 (84.6)	589 (93.8)	6,152 (85.4)
Spanish	802 (12.2)	19 (3.0)	821 (11.4)
Other	206 (3.1)	20 (3.2)	226 (3.1)
Unknown	4 (0.1)	-	4 (0.1)
Area deprivation index quartile
Quartile 1 (least disadvantaged neighborhood)	1,832 (27.9)	127 (20.2)	1,959 (27.2)
Quartile 2	1,350 (20.5)	159 (25.3)	1,509 (20.9)
Quartile 3	1,176 (17.9)	125 (19.9)	1,301 (18.1)
Quartile 4 (most disadvantaged neighborhood)	1,031 (15.7)	97 (15.4)	1,128 (15.7)
Unknown	1,186 (18.0)	120 (19.1)	1,306 (18.1)
Insurance
Private	3,459 (52.6)	433 (68.9)	3,892 (54)
Medicare	2,510 (38.2)	182 (29.0)	2,692 (37.4)
Medicaid	595 (9.0)	13 (2.1)	608 (8.4)
Self-pay	9 (0.1)	—	9 (0.1)
Other	2 (0.03)	—	2 (0.03)

^*Patients with at least one screening mammography encounter in the pre- and postimplementation period who used only traditional scheduling in postimplementation period.

^†Patients with at least one screening mammography encounter in the pre- and postimplementation period who used only self-scheduling in postimplementation period.

The volume of scheduled screening mammography increased from 406.1 per month before the implementation of online self-scheduling to 502 per month afterward. The completion rate of screening mammography increased from 78.9% (320.5 completed encounters per month) to 79.8% (400.6 completed encounter per month).

The unadjusted screening mammography completion rates are shown in Table 2. Using difference-in-difference estimator and adjusting for implementation time, the completion rates in the self-scheduling cohort was 8.4 percentage point (95% confidence intervals, 5.2–11.6) higher than traditional scheduling. The change in screening mammography completion rate from the post- to pre-implementation period was +8.5 percentage point (88.1% postimplementation versus 79.6% pre-implementation) for the self-scheduling cohort and +0.1 percentage point (80.8% postimplementation versus 80.7% pre-implementation) for the traditional scheduling cohort. There was no significant difference in screening mammography completion in the pre-implementation period between the two cohorts (Table 2 and Fig. 1). These findings remained unchanged in analyses adjusted for sociodemographic factors with the self-scheduling associated with a 8.3% increase in completion rate compared with traditional scheduling (Table 2).

Table 2.

Screening mammography completion rates among patients with at least one encounter in pre- and postimplementation period

Statistical Model	Self-Scheduling Cohort* Predicted Margins†	Traditional Scheduling Cohort‡ Predicted Margins†	Difference in Predicted Margins (ie, Difference in Completion Rate)
Unadjusted GEE model	0.833 (0.816, 0.851)	0.807 (0.802, 0.812)	0.026 (0.008, 0.044)
Adjusted GEE model based on implementation time
Pre-implementation period	0.796 (0.772, 0.820)	0.807 (0.800, 0.813)	−0.010 (−0.035, 0.015)
Postimplementation period	0.881 (0.860, 0.903)	0.808 (0.801, 0.814)	0.074 (0.051, 0.096)
Difference in predicted margins (ie, difference in completion rate)	0.085 (0.055, 0.115)	0.001 (−0.008, 0.010)	0.084 (0.052, 0.116)∥
Adjusted GEE model based on implementation time and sociodemographic§
Pre-implementation period	0.799 (0.775, 0.822)	0.8063 (0.800, 0.813)	−0.008 (−0.032, 0.017)
Postimplementation period	0.881 (0.860, 0.902)	0.8058 (0.799, 0.813)	0.075 (0.053, 0.098)
Difference in predicted margins (ie, difference in completion rate)	0.0826 (0.052, 0.113)	−0.0005 (−0.010, 0.009)	0.0831 (0.052, 0.114)∥

GEE = generalized estimating equations.

^*Those who only used self-scheduling after implementation (n = 628 patients).

^†Those who only used traditional scheduling after implementation (n = 6,575 patients).

^‡The average predicted probability of completion of screening mammography.

^§Adjusted for age at examination, race, ethnicity, insurance, and area deprivation index.

^∥Reflects difference-difference estimate.

Fig. 1.

Adjusted screening mammography (SM) completion rate based on implementation time among patients with at least one encounter in pre- and postimplementation period. Self-scheduling cohort: Those who only used self-scheduling after implementation (n = 628 patients). Traditional scheduling cohort: Those who only used traditional scheduling after implementation (n = 6,575 patients).

The e-only Supplementary Table 1 and Figure 2 show the difference-in-difference estimates of screening mammography completion between the self-scheduling and traditional scheduling cohorts over time, based on each of the patients’ sociodemographic variables separately. Self-scheduling patients who were White or Black, non-Hispanic and had Medicare or private insurance had higher screening mammography completion rates compared with traditional scheduling patients. Further, patients in all age groups and neighborhood socioeconomic status had improved screening mammography completion rate with self-scheduling compared with traditional scheduling. However, the difference in screening mammography completion between the two scheduling pathways was not significantly greater when comparing difference age groups, races, ethnicities, neighborhood socio-economic statuses, or insurance types.

Fig. 2.

Difference-in-difference estimates and their 95% confidence intervals for screening mammography completion rates between self-scheduling (those who only used self-scheduling after implementation; n = 628 patients) and traditional scheduling (those who only used traditional scheduling after implementation; n = 6,575 patients) cohorts for selected characteristic variables. ADI NATRANK = area deprivation index national rank; lowest ADI = least disadvantaged neighborhoods.

DISCUSSION

Our retrospective analysis of 7,203 patients with 29,893 scheduled screening mammography encounters before and after implementation of self-scheduling demonstrated an increase in screening mammography completion rates after implementation of online self-scheduling. Further, those who were using self-scheduling as the only scheduling method after its implementation were more likely to complete their screening mammography compared with those who used traditional scheduling.

Self-scheduling screening mammography through a patient portal aims to address patient-level barriers such as calling an imaging center during business hours and waiting on the telephone to reach a scheduler [17,18]. If successful, it offers scheduling flexibility and cost-effectiveness with potential scalability [15] compared with resource-intensive interventions such as mailed or telephone reminders [19,20], educational materials [21], and patient navigation [22]. However, the efficacy of self-scheduling has been debated.

Comparing pre- and postimplementation data, Wadell et al reported that screening mammography completion rate within 6 month of patients’ primary care clinic increased from 22.2% in the pre-implementation period to 49.7% in the post-implementation period. There was significantly more increase among patients with active EMR portal versus inactive EMR portal. Although active EMR portal was used as a proxy for self-scheduling, the study did not report whether all patients with active EMR used self-scheduling [15]. Another study found that self-scheduled screening mammography patients were less likely to cancel or no-show their encounter (0.14% versus 0.22%), although the rates of cancellation or no-show in this study was low (<1%) [9]. In contrast, North et al reported that screening mammography self-scheduled patients had higher no-show encounters (5.7%) compared with staff-scheduled patients (4.6%) [11]. Of note, this study did not break down cancellation rates by scheduling methods [11]. Finally, Sadeghi et al reported 35.3% cancellation or no-show encounters for screening mammography self-scheduling patients versus 25.3% for traditional staff scheduling [12].

Our study results are concordant with prior studies. Although there have been steady increase in screening mammography completion rates at the national level with completion rates reported at 71.5% in 2018, 78.5% in 2019, and 74% in 2021 [23], the difference in completion rates between self-scheduling and traditional scheduling cohorts in our study are likely the result of scheduling method. Our results are similar to Wadell et al [15], and we hypothesize this improvement is likely due to the benefits associated with the convenience of scheduling appointments at patients’ preferred times. Our results are different than Sadeghi et al [12] and North et al [11], who reported increased cancellation and no-shows associated with self-scheduled appointments in the postimplementation period likely due to difference in study designs. In both our study and that of Wadell et al [15], screening mammography completion rates are assessed in the same patient cohorts during the pre- versus postimplementation period of self-scheduling, which would account for patient characteristics that may impact cancellations and no-shows. Lastly, the higher screening mammography completion rate seen in this study compared with national rates is likely due to inclusion of patients with more than two scheduled screening mammography, placing them at a higher adherence rate.

In the current study, 8.7% of patients used self-scheduling as the only scheduling pathway in the postimplementation period. However, it is important to note that among all scheduled screening mammography encounters during the postimplementation period (regardless of patients having an encounter in the pre-implementation period or patients only using self-scheduling during the postimplementation period), 15.6% of encounters use self-scheduling in our health system. Given not every patient has portal access, the rate of self-scheduling adoption is even higher among portal users, which is consistent with prior literature [11,12,14].

Prior studies have reported certain patient groups such as those younger and of non-Black racial background are more likely to use self-scheduling [9,10,12], which is likely reflective of digital divide present among patient portal users [24]. There are controversial data on Hispanics versus non-Hispanics and those living in most versus least disadvantaged neighborhood use of self-scheduling [9,10,12]. In analysis of each patient characteristic separately, our study showed improved screening mammography completion in self-scheduling patients who were White, Black, non-Hispanic, and patients with either Medicare or private insurance, but this improvement was not significantly different across different ages, races, ethnicities, neighborhood socioeconomic status, and insurance types, suggesting that self-scheduling does not widen health disparity.

Many studies demonstrate the impact of online patient portals for increasing patient awareness in preventative screening. Online self-scheduling increases patient autonomy [8], efficiency for both patients and schedulers [25], as well as provides benefits to health care providers [26]. Self-scheduling is just one approach to address the challenge of improving access to screening mammography. If proven effective, it can be implemented relatively quickly without requiring significant institutional changes, such as adjustments in staffing, unlike other resource-intensive interventions. Despite these advantages, barriers to using online patient portals persist, particularly among individuals with low technology and health care literacy. To mitigate language barriers, our patient portals offer modules in Spanish and several other languages. Patient education on how to use portals to self-schedule may further help improve patients’ utilization of self-scheduling. Nevertheless, barriers to completing a screening mammogram remain, reflecting ongoing challenges in access.

Our study had several limitations, one of which included secondary data analysis in which data availability was limited. Moreover, patient factors such as education level, employment status, family history of breast cancer, and access to transportation that can potentially impact screening mammography completion were not available. However, we expect many of these factors to remain constant given our study evaluates changes in screening mammography completion in the post- versus pre-implementation periods for the same cohort of patients. Lastly, given the retrospective nature of our study, it can only suggest general associations. Our study only focused on patients who used only one method of scheduling during postimplementation period. Hence, we did not assess screening mammography completion rate among patients who alternate between scheduling methods during postimplementation period.

TAKE-HOME MESSAGE.

• Self-scheduling was associated with increased screening mammography completion than traditional scheduling among patients with screening mammography encounters pre- and postimplementation.

• Although this increase was more prominent in certain demographics, the degree of increase was not significantly different across ages, races, ethnicities, insurance types, or neighborhood socioeconomic status, suggesting that self-scheduling did not contribute to worsening health disparity gap.

• Patient education on how to use portals to self-schedule may further help improve patients’ utilization of self-scheduling.