Abstract
Background: Preventive care improves patient health and is cost-effective, yet many patients are not up to date on recommended screenings.
Objective: Evaluate the effectiveness of an automated system for outreach to patients in need of annual preventive examinations, cervical cancer screening, and diabetes monitoring labs.
Methods: As part of a quality improvement project, we created a population health algorithm and outreach system which was designed to send e-mail and smartphone notifications to patients overdue for preventive services. The study was a cohort study, with a matched control sample. We compared completion of preventive exams and screenings between the 2 groups, in the 4 weeks following the outreach.
Results: For annual preventive visits, the intervention group had 9.0% more visits (95%CI: 8.2 to 9.7) than the control group. For cervical cancer screening, the intervention group had 3.2% (95%CI: 2.0% - 4.4%) more visits. Lab action orders for diabetes showed the largest increases. The intervention group had 5.2% (2.5% - 7.9%) more patients get bloodwork and 20.8% (16.9% - 24.6%) get more urine microalbumin tests.
Conclusions: A population health outreach system that used reminders for prevention resulted in patients completing appointments for necessary medical services. Such a system, when deployed more broadly could help close care gaps and improve health for people that are asymptomatic but are due for preventive screenings.
- Cancer Screening
- Cohort Studies
- Diabetes Mellitus
- Population Health
- Preventive Health Services
- Primary Health Care
- Quality Improvement
- Text Messaging
Introduction
Screening and preventive care can improve quality of life, improve population health, and are cost-effective.1 An analysis of priorities for prevention found that screening and intervention for many unhealthy behaviors (eg, tobacco, alcohol, unhealthy diet) and cancers as top priorities.1 Despite the importance of screening, estimates indicate that 12 to 18% of women have not had recent cervical cancer screening and 25 to 50% are not up to date with current guidelines; more than 50% of cancers are in women who are not up to date in screening.2⇓–4
Barriers to provision of preventive exams and screening tests are multifactorial. These factors include systems, clinicians, and patients. Most preventive care screenings are delivered by primary care clinicians working in systems with insufficient support for care outside of the office visit. Electronic health records (EHRs) typically provide point of care decision support (eg, pop-up reminders) during a visit, although these may be subject to alert fatigue.5 In most fee-for-service insurance models, preventive services are covered and typically occur during a general health check or “annual physical examination.” Research shows that patients who receive these general health checks have improvement in patient-reported outcomes and greater uptake of clinical preventive services.6 Outside of these annual exams, clinicians with busy schedules and a short appointment length may not have sufficient time to perform preventive coordination.7
Patient-driven factors also affect screening compliance. While people value their health, they may not engage in wellness exams or screenings for various logistic reasons. For screenings such as colorectal and cervical cancer, there are many factors why patients do not seek care. These typically include knowledge, perceived efficacy, perceived risk, and social stigma.8 In addition, since prevention targets asymptomatic disease, there is no prompt or symptom that reminds patients to seek care.
Primary care practices can overcome the above barriers by creating models that allow more time for preventive coordination through longer office visits and virtual care. Practices can also lower barriers to access and increase engagement with next-day appointment availability and access to care via remote technology.
Another approach to improving preventive care is to increase patient engagement with reminders for care. Reminders, in the behavioral economic literature, are a type of nudge that can intentionally and transparently prevent irrational decisions (ie, not seeking preventive care).9 Many health systems rely on reminders in the electronic health record (EHR) that target clinicians.10 In contrast, reminders to patients have the potential to promote patient engagement in their care. Reminders can be deployed via phone calls, text messages, e-mails, mailings, or smartphones.11 However, few studies have looked at the effects of modern types of reminders, such as push-notifications in smartphone applications.
The objective of this quality improvement study was to develop and evaluate effectiveness of an automated system for outreach to patients in need of annual preventive examinations; cervical cancer screening; and labs to prevent the complications of diabetes mellitus.
Methods
Setting
This study was conducted at One Medical, a membership-based primary care practice based in urban and suburban locations throughout the United States. At the time of implementation, the practice served mostly a commercially-insured population. As previously described, the practice is an adaptation of a Patient-Centered Medical Home (PCMH), as it is built on the core attributes of primary care, along with enhanced access, a quality-improvement structure, and some blended payments.12 The model has been previously shown to decrease costs and improve quality of care for chronic disease.13,14
Development of Clinical Algorithms
The goal of our quality improvement team was to create an automated reminder system that would notify patients when they were due for preventive services. The first step in creating our system was to specify clinical logic for the outreach. We selected a subset of conditions that have well-established clinical follow-up periods and/or quality metrics: annual preventive exams, cervical cancer screening visits, and diabetes-monitoring labs. These algorithms were developed for adults meeting the inclusion/exclusion criteria for each condition (Table 1). A group of clinical leaders developed the algorithms, which were then coded by data scientists in structured query language and executed to retrieve eligible patient lists from our clinical data repository.
Action Item Inclusion and Exclusion Criteria
One Medical member patients who had not received an annual preventive examination in the past 12 months were selected to schedule an annual preventive care visit. The purpose of the annual preventive care visit is to focus on the patient’s physical and mental health priorities, screening for unhealthy behaviors, and administering clinical preventive services. At the time we started this work in 2021, 23% of our patients had not had an annual examination in over a year.
For cervical cancer screening, we selected patients who had an annual physical examination in the past 12 months but did not have documentation of cervical cancer screening. We used these criteria to select patients that were engaged with care and likely had a primary care clinician who recommended a timeline for cervical cancer screening. Approximately two-thirds of our patients had seen us in the past year, but at the time of our intervention, were due for screening or documentation. The clinical algorithm called for these patients to book a visit for cervical cancer screening. However, some of these patients may have had screenings at other practices, so the outreach would remind the patients to discuss their need for testing with their primary care clinician.
For diabetes, we used standard clinical guidelines that state that patients should have a Hemoglobin A1c (HbA1c) every year, or every 3 months if they have an HbA1c >8.0; and a urine microalbumin (UMA) every year. The clinical algorithm called for these patients to have labs drawn. At the time of intervention, 59% of patients were compliant on having an up-to-date A1c, and 74% were compliant on having a recent UMA.
Leveraging Technology
Our proprietary EHR is integrated with an iOS and Android app, which allows for push notifications. For those without the mobile app, the EHR can send notifications to an e-mail address. We used this notification feature to build a new feature called “Action Items.” (Figure 1) These Action Items notified patients to either book a visit or come to the lab for a necessary test. In the case of a visit Action Item, the app would suggest appointment slots to book to complete this action. (Figure 2) For the lab action items, the action would indicate the closest lab and hours of operation. (Figure 3) The orders for the labs were automatically ordered following an approved clinical algorithm, overseen by Medical Director of Population Health. When receiving the action items, patients also had the option to state they already completed it, or to “decline” the action item.
Screenshots of notification.
Appointment booking.
Lab action item.
We additionally created a system that would allow these orders to be sent out in batches. A SQL query would pull a list of patients from our patient data repository and send the list to our “Action Engine” in our EHR. The Action Engine, before sending out the Action Items, factors in regional demand and capacity for visits to balance access for patients needing both preventive and acute care.
Study Design
These new features were rolled out as a quality improvement initiative to improve preventive care in our health system in June of 2021. Thus, we undertook an evaluation after the implementation of the features, using patients who did not receive the action items as a control group.
For the annual preventive examination and cervical cancer screening, the patients who received the action items were a random sample of patients in each Metropolitan Statistical Area (MSA). In each MSA we capped the number of action items we sent out, to make sure there would enough appointment slots available for each patient who received an action item. We considered the patients who received the action items in the “intervention” group. The patients who did not get an action item in the initial round of outreach were eligible for our matched-control population. After the implementation was complete, we created a matched sample to the intervention group from the population that did not receive an action. We matched this sample without replacement on location (by MSA), gender, and age.
For the diabetes lab orders, the quality improvement team decided prospectively to create a holdout group of 30% of the patients, so they could evaluate the effects of the initial outreach. Thus, the intervention group was a random selection of patients due for labs in each MSA.
To assess the effectiveness of this new tool in promoting clinical engagement our primary outcome was the rate of appointments booked or labs completed within 4 weeks, compared with a control group. We compared the proportion of patients who booked an appointment or received a lab test in the 4 weeks following the outreach, using z-tests of proportions. In addition, for the cervical cancer screening action items, we also analyzed the rate of receiving a completed lab result for the screen within 4 weeks. All differences are reported with 95% confidence intervals, using Python for analysis.
Results
Table 2 shows the demographic characteristics of the intervention and control patients. The average age for the preventive visits and cervical cancer screenings was young, with an average age of 37 to 38. The average age for the diabetes samples were older, with a range of 47 to 51. The table shows that our statistical matching created control samples similar to our intervention samples.
Demographics for the Intervention and Control Groups
Rates of Completion of Visits and Labs Four Weeks after Action Items Were Sent Out
For preventive visits, the intervention group had 9.0% more visits (95%CI: 8.2 to 9.7) than the control group (Table 3). Cervical cancer screening action items were also effective, with the intervention group having 3.2% (95%CI: 2.0%-34.4%) more visits. To look at this effect on screening rates, we looked at cervical screening results. The intervention group had a rate of 1.9% of receiving labs, while the control cohort had a rate of 1.0% (95%CI on difference: 0.42%-1.4%). Lab action orders for diabetes showed the largest increases. For A1c, the intervention group had 9.0% of patients come in for labs, compared with only 3.8% for the control group (5.2% difference, 95%CI: 2.5 to 7.9). Results were even larger for UMA, with a 20.8% (95%CI: 16.9 to 24.6) difference between intervention and control.
Discussion
This study describes the application of population health methods powered by clinical algorithms and advanced technology in a primary care setting. We demonstrated statistically and clinically meaningful improvements in preventive care services in preventive care visits, cervical cancer screenings, and diabetes care lab tests.
The NASEM Implementing High Quality Primary Care report highlights the potential role of digital health tools in improving population health, but notes that many of these tools are currently limited to EHR functions at the point of care.10 As a result, they often do not truly address proactive population health management. While the report mentions patient registries as effective tools for proactive outreach, it does not provide specific examples or cite studies that combine registries with other digital approaches, such as patient portals. In contrast, the system described in this article goes beyond the point-of-care model by creating an automated patient registry that generates proactive notifications within an EHR-linked smartphone application. This novel integration represents a step toward the type of automated, proactive care system that the NASEM report calls for.
We developed this unique approach to population health integrated in primary care by addressing key barriers to preventive health screenings. Many people have deferred essential care during the pandemic and advanced primary care models such as described here will be critical to close these care gaps and prevent future burden of illness. We described how we built and implemented a system to determine which patients are due for engagement with their primary care clinician and reach out to these patients to prompt a preventive health action. This system, which works through a new type of outreach called an “Action Item,” utilizes both e-mail alerts and smartphone notifications to close gaps in care.
Closing screening gaps for diabetes are an ideal use of this type of system. Testing HbA1c and UMA are a core part of diabetes management and quality metrics. While the raw numbers of patients with diabetes due for HbA1c were small in this study sample, we saw large increases in engagement for those that were due. Since these labs are sent to the primary care clinician, the clinician can either make sure their diabetes is under control or message the patient to follow-up for care. While this analysis does include the values of A1c and downstream actions, many of these patients likely received interventions from their clinician designed to better control their diabetes.
Most previous studies in this area have looked at sending reminders for appointments already booked. One study sent text messages to patients with diabetes to remind them of appointments and sent physicians a notification for abnormal results. It found greater appointment adherence and improvements in biometrics in the intervention.15 Several other studies in a systematic review show that reminders, usually in the form of text messages, improve adherence rates to appointments.16 Thus, the theory of reminders for care is plausible, although our intervention sought to engage patients before they booked an appointment.
Studies on cervical cancer screening have looked at reminders to book an appointment. These studies have shown sending letters to women due for screening improves cervical cancer screening rates.17,18 Text messages to get a mammogram seem to add benefit to those who receive outreach with letters.19 According to a systematic review, text messages do increase rates of screening for breast, cervical, colorectal, and lung cancers.20
Other studies have tried to improve adherence to care using tools in the office setting. One study found that providing patient education in addition to a clinician app, did not improve cervical cancer screening rates.21 A study that prompted patients to select which preventive items they wanted to complete in a patient portal, before an upcoming visit, improved influenza and mammography rates, but not cervical cancer or bone density screening rates.22
The main limitation of our study was the lack of a prospective randomized design, which was impractical due to the constraints on a project within a learning health system. While a prospective randomization at the regional level would be a stronger design, a matched cohort study is adequate for this type of implementation study. In addition, our study was conducted in a predominantly commercially insured population living in major MSAs, with high rates of smartphone usage. This limits the generalizability of our findings, particularly for populations that are underinsured or uninsured, or for those in rural areas with limited access to digital tools. Racial, ethnic, and socioeconomic disparities in health care access—particularly among underserved populations—are well-documented and often result in reduced access to preventive services and lower engagement with health care interventions. Future studies should focus on implementing similar outreach programs in more diverse populations, including those who face greater barriers to care, such as lower income groups, racial and ethnic minorities, and patients without reliable access to smartphones or digital technology. The study was conducted during 2021, a period when the COVID-19 pandemic significantly impacted health care access, including preventive services that required in-person visits, such as cervical cancer screenings. This disruption likely affected the uptake of preventive services in our study population, potentially contributing to lower rates of completed visits compared with what would be expected in a normal year. Future research should further explore how digital health interventions can help mitigate these disruptions, particularly in the context of pandemic-related challenges to health care access.23
The NASEM Implementing High Quality Primary Care report calls for EHR systems to “make care more proactive” and “automate more care.” This article presents a real-world example of such a system in action. We found that a population health outreach system for prevention resulted in patients completing appointments for necessary medical services. Such a system, when deployed more broadly could help close care gaps and improve health for people that are asymptomatic but are due for screenings. Future studies should look at whether outreach systems result in improvements in disease outcome measures, and what types of patients are most responsive to this type of outreach.
Acknowledgments
Arielle Slam and Colleen Bouey contributed to the design and earlier versions of this manuscript.
Notes
This article was externally peer reviewed.
Funding: None.
Conflict of interest: None.
To see this article online, please go to: http://jabfm.org/content/38/2/239.full.
- Received for publication February 23, 2024.
- Revision received September 25, 2024.
- Accepted for publication October 7, 2024.
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.