Feasibility of a mobile application to track longitudinal pain outcomes after epidural joint injections☆ Authors: Patricia Z Zheng, Ted Miclau, Brandon Goenawan, Gabriel Johnson, Conor O'Neill, Lyndly Tamura, Peter I-Kung Wu, Matthew Smuck Published: 2025 ABSTRACT AbstractBackgroundTracking spine intervention outcomes is costly and time-consuming, hindering research. Mobile health (mhealth) applications offer a promising solution to reduce the need to hire personnel to recruit patients and collect data. However, evidence supporting the use of mHealth to study pain and spine interventions is limited and heterogeneous.ObjectivesThis study aims to understand how recruitment methods influence patient recruitment and retention in a digital study tracking pain. We hypothesize that in-person recruitment, especially with a physician rather than a research assistant, leads to higher recruitment and retention.MethodsWe designed a prospective observational study of patients scheduled for spine injections at an academic outpatient surgical center. Eligible patients were recruited via three methods: in-person research assistant, remote research assistant via phone, or in-person clinician. Upon enrollment, a mHealth application prompted patients to complete post-injection surveys at pre-defined intervals based on injection type. Enrollment and retention rates were then compared across recruitment groups.ResultsOf 194 patients approached, enrollment rates differed significantly by recruitment strategy: 94.1 % for in-person physicians (N = 32 enrolled/34 approached), 85.0 % for in-person research assistants (N = 51/60), and 30.0 % for remote research assistants (N = 30/100) (p < 0.009, Fisher's exact test). Of those recruited, survey completion rates were 74.4 % (physician-recruited patients), 55.1 % (in-person research assistant), and 58.4 % (remote research assistant). Neither demographic factors (age, sex, employment, payer type) nor injection type influenced enrollment. Retention differed significantly across groups (p = 0.0174, Kruskal-Wallis test), with physician-recruited patients completing the most surveys. Moreover, Medicare patients had significantly higher completion rates than those with Medicaid.ConclusionTracking post-injection pain using mHealth applications is feasible, though success depends heavily on recruitment strategy. In-person physician-led enrollment resulted in the highest recruitment and retention rates though the drop-out rates were still significant. Future studies should focus on refining recruitment approaches to optimize patient participation. Introduction Tracking outcomes after pain and spine interventions is time-consuming and costly, which limits the scope and pace of spine intervention research. One recent randomized controlled study examining the efficacy of lumbar epidural steroid injections for lumbar spinal stenosis required a budget of more than two million dollars [1]. Studies have shown that much of this research budget is spent on hiring personnel needed to contact patients to collect outcome measures [2,3]. Mobile health (mHealth) applications offer a promising solution to these challenges as they provide direct access to patients, thereby reducing reliance on research personnel and increasing the ease and cost-effectiveness of the research process [4,5]. Furthermore, 91 % of Americans now own smartphones with internet connections, making mHealth applications a more generalizable solution to conducting research [6,7]. mHealth applications can transform the way longitudinal clinical information is collected. They prompt patients to report outcomes (allowing for greater reporting frequency with reduced burden on patients), provide a personalized research process (by eliminating redundant or irrelevant questions based on prior responses), and potentially reduce recall bias (as patients are queried about their pain in their normal living environments) [8,9]. Potential barriers to the use of mHealth in clinical studies of pain and spine interventions include poor recruitment and retention, given the lack of a “human touch.” Virtual recruitment for studies in general is poor when compared to in-person recruitment [[10], [11], [12]]. Similarly, a review of 21 papers using mHealth apps for research found that retention rates ranged as low as 10.4 % [11]. A mHealth application based digital care pathway for patients undergoing spine surgery had low patient usage rates in the post-operative setting, with only 30 % of patients completing at least one survey [13]. Long term engagement is especially challenging. In a large cardiovascular mHealth study, approximately 71 % of participants completely disengaged within 3 months [14]. An analysis of 100,000 participants in digital health studies found that more than half of the participants discontinued participation within the first week of a study [15]. Recruitment and retention methods can be enhanced by in-person engagement, especially involving clinicians. Research found that patients were more amenable to joining a study if they were approached directly by a physician, specifically their usual doctor [16]. In another instance, patients with cancer were more likely to participate in an RCT study if the physician who was involved in their care was trustworthy, enthusiastic, and a good communicator [17]. Consequently, the type of recruitment method employed (i.e. physician-led versus clinical research staff-led) may impact the recruitment and retention of patients in digital-based studies that track pain. Specific to mHealth apps, Pratap and colleagues found that having a clinician introduce the mHealth application is one of the 4 factors significantly associated with increased retention, increasing the median retention time by 40 days [15]. Given these considerations, the aim of this paper is to assess the feasibility of a custom mHealth application created to track longitudinal pain outcomes after various spine injections. Specifically, we tested the recruitment and retention capabilities using various recruitment strategies, including using clinician recruitment. However, as clinician time is often limited in a busy interventional pain and spine practice, we compared the effectiveness of clinician recruitment versus research assistant recruitment, both in-person and remote. Methods Study design This was a prospective observational study conducted at a single academic outpatient surgical center. Approval for all study procedures was obtained from the Institutional Review Board (IRB). Patient selection Patients scheduled to undergo any of the following spinal injections were identified from the electronic medical record (EMR) and screened for study recruitment: epidural steroid injection (ESI), joint injection, medial branch block (MBB), and medial branch radiofrequency ablation (RFA). Inclusion criteria were comprised of the following: (1) ≥18 years old with medical decision-making capacity, (2) ownership of a phone or computer with internet connectivity. Patients were excluded if they had: (1) difficulties providing consent in English, (2) ongoing litigation involvement from an injury, or (3) a condition or pregnancy making spine interventions unsafe. Participation was optional and patients were informed that participation in the study would not impact the care that they received. They were also informed that they would not receive any compensation for participating. Recruitment and enrollment process After screening for inclusion and exclusion criteria, qualifying participants were approached for study recruitment through one of three modalities: in-person research assistant, remote research assistant, or in-person physician recruitment. The modality used for consecutive patients on a given procedure day was determined pragmatically given the availability of the physician and the research assistant. During the recruitment process, informational sheets were distributed to patients that outlined study details, including how to access and use our web-based mobile application. Patients in the in-person group were approached on the day of their injection in the pre-operative area of the surgical center and provided paper copies of the informational handout and consent forms. Patients in the remote recruitment group were called by the research assistant the day before their injection, no more than two times at approximately 12 p.m. and/or 5 p.m., and the research assistant followed a phone script for recruitment. If interested, they were then emailed a DocuSign consent forms to sign electronically. Interested patients who provided verbal consent and signed HIPAA authorization forms were officially enrolled in the study. Data collection Upon enrollment, the respective recruiting staff member noted the patient's mobile phone number, date, and type of scheduled injection. This information was entered into a custom-built web-based mHealth application, which then texted a link to enrolled patients to respond to post-injection surveys. Electronic survey measures included average hourly and daily pain, maximum daily pain, the Patient-Reported Outcomes Measurement Information System (PROMIS)-29 survey, and the Patient Global Impression of Change (PGIC) survey. PROMIS-29 is a reliable and valid tool to assess pain intensity using a single 0–10 numeric rating item within seven health domains (physical function, fatigue, pain interference, depressive symptoms, anxiety, ability to participate in social roles and activities, and sleep disturbance) [18]. Similarly, PGIC is a short, widely used questionnaire that serves as an indicator of overall changes in pain over time, is a clinically relevant tool to assess perceived impact of disease management (e.g., fibromyalgia) [19], and may be important to consider among patients living with chronic pain [20]. The exact measurement outcomes and frequency of requested surveys depended on the type of injection performed. For ESI and joint injections, outcomes were average daily pain and worst daily pain on the day prior to the injection, current pain hourly for 8 h on the day of injection, average and worst daily pain for 14 days, and average and worst daily pain collected monthly for 3 months after the day of injection, along with PROMIS 29 prior to the injection and repeated at 1 month and at 3 months adding the PGIC, all of which totaled 26 surveys. For MBBs, prompted outcomes were average daily pain and worst daily pain on the day prior to injection as well as a current pain assessment hourly for 8 h after the injection, which totaled 9 surveys. For medial branch RFA, prompted outcomes were average daily pain and worst daily pain on the day prior to injection, current pain hourly for 8 h on the day of injection, average and worst daily pain for 14 days, and average and worst daily pain collected monthly for 6 months after the day of injection, along with PROMIS 29 on the day prior to the injection and repeated at 3 month and 6 months adding the PGIC, all of which totaled to 45 surveys. Sociodemographic information such as age, sex, insurance status (Medicare, Medicaid, Private, Workers Comp), and employment status (employed, part time, not employed, disabled, retired, and unknown) were also recorded. Data handling complied with UCSF Institutional Review Board (IRB) requirements. All mobile device data was collected via the self-designed application with data stored on the Amazon Web Services server. This only included de-identified data except for the patients’ phone numbers. All individual patient data was identified by their phone number only, and all identifiable medical data was stored on REDCap. Primary outcomes Our study's primary outcomes were (1) recruitment, defined as the percentage of patients successfully enrolled in the study from the total number targeted, and (2) retention, defined as the percentage of requested pain surveys completed by enrolled patients. These two measures were then compared between the different branches of patient recruitment (i.e., physician versus research assistant; remote versus in-person). At times, the mobile application malfunctioned given changes in security requirements. Consequently, retention rates were measured as percentage of surveys completed rather than percentage of surveys sent out. If patients declined to participate or if they later withdrew, they were encouraged to share their reasons, which were recorded. Statistical analysis All statistical analyses were conducted with R version 4.2.2 using the following packages: lubridate (version 1.9.2; Grolemund G, Wickham H, 2011), report (version 0.5.7; Makowski D et al., 2023), MatchIt (version 3.0.2; Ho D et al., 2011), gtsummary (version 1.7.2; Sjoberg D et al., 2021), ggplot2 (version 3.4.1; Wickham H, 2016), readxl (version 1.4.2; Wickham H, Bryan J, 2023), zoo (version 1.8–12; Zeileis A, Grothendieck G, 2005), and tableone (version 0.13.2; Yoshida K, Bartel A, 2022). All figures were generated through Microsoft Excel. Differences in enrollment success rates between recruitment groups were assessed for statistical imbalance using a Fisher Exact test and logistic regression. A Kruskal-Wallis Test and a linear regression analysis were conducted to assess differences in retention rates between recruitment groups. Results 194 participants were approached for recruitment and had a mean age of 59.9 years (SD: 16.04 years), with 41.2 % being male. In terms of insurance status, 75 patients had Medicare, 23 had Medicaid, 94 were privately insured, and two patients claimed Workers’ Compensation. Majority of patients were either fully employed or retired. Table 1 displays the demographics of all patients approached, organized by recruitment method. Although patients recruited by the remote assistant were slightly older than the in-person physician-recruited groups, variables like sex, employment status, payer status, and injection type did not differ significantly among the different recruitment groups. Table 1Demographic Characteristics of approached Patients by Recruitment Type (N = 194)Continuous variables are presented as mean (SD) and analyzed using the Kruskal-Wallis Test; categorical variables are presented as n (%) and analyzed using either a Fisher's Exact Test or a Chi-squared Test.Table 1VariableIn-Person MD Group (N = 34)In-Person RA Group (N = 60)Remote RA Group (N = 100)p-valueAge (years), mean (SD)54.4 (14.2)6b.0 (15.8)61.7 (16.5)0.074cGender0.516b Male17 (50.0)24 (40.0)39 (39.0) Female17 (50.0)36 (60.0)61 (61.0)Employment Status0.315a Employed19 (55.9)28 (46.7)37 (37.0) Part-time2 (5.9)2 (3.3)2 (2.0) Not Employed4 (11.8)6 (10.0)7 (7.0) Disabled0 (0.0)2 (3.30)9 (9.0) Retired7 (20.6)18 (30.0)34 (34.0) Unknown2 (5.9)4 (6.7)11 (11.0)Payer Status0.116a Medicare8 (23.5)26 (43.4)41 (41.0) Medicaid2 (5.9)6 (10.0)15 (15.) Private23 (67.6)28 (46.7)43 (43.0) Workers Comp1 (2.9)0 (0.0)1 (1.0)Injection Type0.354a ESI17 (50.0)33 (55.0)57 (57.0) Joint Injection4 (11.8)11 (18.3)23 (23.0) MBB10 (29.4)9 (15.0)15 (15.0) RFA3 (8.8)7 (11.7)5 (5.0)aFisher's Exact Test.bChi-squared Test.cKruskal-Wallis Test. Demographic Characteristics of approached Patients by Recruitment Type (N = 194) Continuous variables are presented as mean (SD) and analyzed using the Kruskal-Wallis Test; categorical variables are presented as n (%) and analyzed using either a Fisher's Exact Test or a Chi-squared Test. Fisher's Exact Test. Chi-squared Test. Kruskal-Wallis Test. Of the 194 participants approached, 113 patients (58.2 %) were successfully recruited to our study (Fig. 1), with Table 2 showing the patient demographic breakdown by recruitment type. Similar to the total number of patients who were approached, there was no significant difference in the age, sex, employment, payer status, or type of injection enrolled patients received among the three types of enrollment methods. Recruitment flow diagram RA = Research Assistant. Table 2Demographic Characteristics of Enrolled Patients by Recruitment Type (N = 113)Continuous variables are presented as mean (SD) and analyzed using the Kruskal-Wallis; categorical variables are presented as n (%) and analyzed using either a Fisher's exact test or a Chi-squared Test.Table 2VariableIn-Person MD Group (N = 32)In-Person RA Group (N = 51)Remote RA Group (N = 30)p-valueAge (years), mean (SD)54.7 (14.6)59.4 (15.5)59.2 (14.9)0.074cGender0.638b Male16 (50.0)23 (45.1)11 (36.7) Female16 (50.0)28 (54.9)19 (63.3)Employment Status0.829a Employed17 (53.1)25 (49.0)14 (46.7) Part-time2 (6.3)2 (3.9)0 (0.0) Not Employed4 (12.5)5 (9.8)2 (6.7) Disabled0 (0.0)2 (3.9)2 (6.7) Retired7 (21.9)14 (27.5)8 (26.7) Unknown2 (6.3)3 (5.9)4 (13.3)Payer Status0.270a Medicare8 (25.0)21 (41.2)11 (36.7) Medicaid2 (6.3)5 (9.8)5 (16.7) Private21 (65.6)25 (49.0)14 (46.7) Workers Comp1 (3.1)0 (0.0)0 (0.0)Injection Type0.341a ESI16 (50.0)30 (58.8)16 (53.3) Joint Injection3 (9.4)9 (17.6)7 (23.3) MBB10 (31.3)5 (9.8)4 (13.3) RFA3 (9.4)7 (13.7)3 (10.0)aFisher's Exact Test.bChi-squared Test.cKruskal-Wallis Test. Demographic Characteristics of Enrolled Patients by Recruitment Type (N = 113) Continuous variables are presented as mean (SD) and analyzed using the Kruskal-Wallis; categorical variables are presented as n (%) and analyzed using either a Fisher's exact test or a Chi-squared Test. Fisher's Exact Test. Chi-squared Test. Kruskal-Wallis Test. In the 34 in-person physician-approached group, two participants were uninterested, leaving 32 (94.1 %) in the final in-person physician-enrolled group. In the 60 in-person research assistant approached group, six participants were uninterested, two patients were interested but had difficulties being enrolled, and one patient was found out subsequently to not have been able to read or speak English well, resulting in 51 (85.0 %) patients being in the final in-person research assistant-enrolled group. For the 100 remote research assistant approached group, 62 patients were contacted, of which twenty-three were uninterested, and another nine were interested but had difficulties accessing the DocuSign consent forms, culminating in 30 (30.0 %) patients in the final remote research assistant enrolled group. Enrollment differed significantly between the three groups (Fisher's exact test, p < 0.009), with remote recruitment exhibiting significantly less success than in-person recruitment (Table 3). Neither injection type nor demographics (age, sex, payer status, or employment) influenced recruitment rates. Table 3Logistic regression modelof predictors on recruitment rates (N = 194).Table 3CharacteristicLog (OR)195 % CI1p-valueRecruitment Type In-Person MD–– In-Person RA-0.57−2.58, 1.030.52 Remote RA−3.60−5.54, −2.23<0.001Age−0.01−0.06, 0.030.48Gender Female–– Male0.07−0.82, 0.950.88Employment Status Disabled–– Employed0.86−0.78, 2.600.32 Part-time0.31−2.28, 3.040.82 Not Employed0.71−1.25, 2.800.48 Retired0.16−1.53, 1.950.86 NA0.60−1.32, 2.610.55Payer Status Medicaid–– Medicare0.01−1.34, 1.390.99 Private−0.50−2.04, 1.040.52 Workers' Comp−1.01−5.77, 3.640.70Injection Type ESI–– Joint Injection−0.21−1.25, 0.820.69 MBB−0.43−1.59, 0.700.46 RFA1.29−0.47, 3.410.18ESI = Epidural Steroid Injection, MBB = Medial Branch block, RFA = Radiofrequency Ablation.1OR = Odds Ratio, CI = Confidence Interval. Logistic regression modelof predictors on recruitment rates (N = 194). ESI = Epidural Steroid Injection, MBB = Medial Branch block, RFA = Radiofrequency Ablation. OR = Odds Ratio, CI = Confidence Interval. Overall, 2589 surveys were sent out, and 1671 surveys were completed (a 64.5 % completion rate). Of the enrolled participants, the percentage of completed surveys was 74.4 %, 55.1 %, and 58.4 % amongst the in-person clinician (N = 32), in-person research assistant (N = 51), and remote research assistant recruitment groups (N = 30), respectively (Fig. 2). Kruskal-Wallis Test demonstrated a significant difference in the percentage of surveys completed by patients between groups (p = 0.017), with the in-person research assistant recruitment group answering significantly fewer surveys than those recruited by a physician (Table 4). There was also a trend (p = 0.054) for decreased retention in the remote research assistant as compared to in-person physician recruitment. Moreover, patients who had Medicare insurance were significantly more likely to have a higher survey response rate than those with Medicaid. Males had slightly lower retention, but not significantly different from females. Outside of recruitment method and insurance status, patients’ age and the type of injection received were not found to impact survey retention. Study enrollment and retention rates between recruitment groups. Table 4Linear regression model of predictors on retention rates (N = 113).Table 4CharacteristicBeta95 % CI1p-valueRecruitment Type In-Person MD–– In-Person RA−0.229−0.384, −0.0740.004 Remote RA−0.175−0.354, 0.0030.054Age−0.001−0.009, 0.0060.742Gender Female–– Male−0.113−0.247, 0.0210.098Employment Status Disabled–– Employed0.051−0.296, 0.3990.769 Part-time−0.124−0.626, 0.3770.625 Not Employed−0.092−0.490, 0.3070.650 Retired0.018−0.355, 0.3910.924 NA−0.191−0.594, 0.2110.348Payer Status Medicaid–– Medicare0.3270.068, 0.5850.014 Private0.217−0.033,0.4680.088 Workers' Comp0.296−0.410, 1.0030.408Injection Type ESI–– Joint Injection0.007−0.173, 0.1880.935 MBB0.026−0.148, 0.2000.767 RFA0.143−0.076, 0.3620.198ESI = Epidural Steroid Injection, MBB = Medial Branch block, RFA = Radiofrequency Ablation.1CI = Confidence Interval. Linear regression model of predictors on retention rates (N = 113). ESI = Epidural Steroid Injection, MBB = Medial Branch block, RFA = Radiofrequency Ablation. CI = Confidence Interval. Discussion This prospective observational study focused on enrollment and retention for a mHealth application designed to track outcomes after spine injections. In total, 194 patients scheduled to receive various spinal injections were approached at a single academic outpatient surgical center. We found that the recruitment method significantly impacted both enrollment and retention. Patient's willingness to enroll in and remain in the study was influenced by who approached them (physician vs. research assistant) and how they were approached (e.g., in person vs. remotely). Enrollment success rates were significantly higher in the in-person recruitment groups (94.1 % for physicians and 85.0 % for research assistants) compared to the remote research assistant recruitment group (30.0 %). Once patients were enrolled in the study, the retention rates, defined as the percentage of surveys completed, was also significantly higher if patients were recruited by a physician. Most of the confounders we tracked did not affect recruitment and retention including: age, gender, employment, injection type. Medicaid insurance was the only confounder that was found to significantly reduce retention. While there is no definitive study pinpointing why Medicaid patients are less likely to remain engaged in research compared to those with other insurances, it is often noted that patients with Medicaid generally have more limited financial resources, which can lead to increased life stressors and a greater focus on pressing daily concerns, making survey participation less of a priority [21,22]. As we did not track income and educational level in our study, Medicaid insurance status may be a surrogate for lower income and educational levels, which have been previously associated with lower participation in clinical studies [23,24]. Future projects can test if the provision of targeted support (e.g. education and/or financial incentives) can address these payer-related disparities in retention. Our study casts doubt on whether mHealth apps can be adequately used to recruit and retain a high percentage of patients to follow outcomes after spine injections. Although there is no universal gold standard retention rate, some studies suggest 80 % is acceptable for most trials, with 90 % retention being considered excellent [25]. In contrast, our study showed a high drop-off rate that ranged from 26 % to 45 %, which is below the suggested 80 % retention rate, raising concerns for attrition bias. While there are few similar prior studies to which we can make a direct comparison, our findings echo the results of Ponder and colleagues' study demonstrating feasibility of their mHealth application for patients undergoing elective spine surgery in a large academic health system [13]. Both their study and our study utilized patient-reported outcome measures and monitored and reported patient participation rates, but the findings demonstrated some differences. Their survey participation rate was less than 30 % in the post-operative setting, whereas our overall participation rates ranged between 55.1 % and 74.4 %. Relative to the Ponder study, our study's participants were slightly older (59.9 years versus 59.0 years) and comprised of more men. In addition, our study compared outcomes between recruitment groups but did not report patient satisfaction, which was one focus of their study. Despite the potential advantages of digital methods in tracking injection outcomes, in-person recruitment for initial study enrollment was found to be the most successful (94.1 % for in-person physician and 85.0 % for in-person research assistant vs 30.0 % for remote research assistant). There are several possible reasons for this. Even though the phone-based caller identification listed our academic institution, patients may have been unavailable, may have thought that the calls were spam. Provided that patients received documents to sign via email after the phone call, there may have been further confusion surrounding how to sign the online forms, or patients may have changed their minds about participating after reading through the study details. Future research may consider enhancing research assistant training, in particular walking patients through the online consent forms to improve remote recruitment rates. In a multisite randomized control trial where RAs were trained through a standardized three-part certification process to learn, practice, and calibrate their interviewing, technical, and clinical skills, the trained research assistants showing enhanced ability to engage with patients [26]. For enrollment and retention, having in-person clinician recruitment was the most successful. Multiple explanations could account for this. Pratap and colleagues theorized that patients’ trust in clinicians may contribute to the higher retention rates in mHealth application based studies recommended by physicians [15]. Patients likely have a stronger personal relationship with their doctor compared to a one-time interaction with a research assistant and thus might feel obliged to participate out of respect. However, clinicians in a pain and spine intervention clinic are often busy and have limited time to talk to patients about study participation. As mHealth applications become increasingly widespread, more research will be needed to understand how to best harness their potential. For example, these applications can help to reduce provider and patient burdens associated with clinical outcomes monitoring and research. They may also allow researchers to collect patient information at higher volumes and reduced costs relative to the traditional research methods conducted fully in person; and they may facilitate patient registries. However, if a large number of patients need to be approached to recruit an adequate sample of patients to join a study, or if recruited patients only answer a small proportion of prompted surveys, then employing a mHealth application may not be a viable or cost-effective method to conduct pain and spine intervention research. Improved understanding of recruitment and compliance will inform how best to design future mHealth applications to incentivize participation and optimize patient and provider satisfaction. Our study is subject to several limitations. First, due to the nature of recruitment, there may be consent bias, in which patients who agree to participate in the study may differ at baseline from patients who opt not to participate. Further, in reporting subjective pain levels, patients may feel pressured to report pain levels either higher or lower than reality, depending on personal interests and/or their perceptions of provider preferences. Third, most of our patients are older and English-speaking, which reflects our institution's demographics of injection patients, but may not be representative of our country's demographic makeup for injection patients. As a result of our academic center's demographic makeup and the fact that this was an initial study the mHealth app feasibility, our consent forms and surveys were only written in English. Next, while we obtained qualitative feedback from those who did not enroll or dropped out of the study, we did not gather qualitative data from those who enrolled and engaged in the study. Thus, we can only speculate about which factors influenced patients to enroll in the study and respond to the multiple surveys. Finally, as our in-person study recruitment occurred in the pre-operative area of our surgical center and the research assistants were not able to join the clinician on every injection day, there was not consecutive recruitment of eligible patients. Despite each of these considerations that affect the interpretation of our findings, our study has several strengths. To our knowledge, no prior work has examined the feasibility of a mHealth application specifically for pain and spine injection longitudinal pain outcomes, nor have many studies reported the enrollment success and participation rates for mHealth application-based studies. Additionally, our study provides insights into both the design of effective mHealth applications and the success rates of various recruitment methods that must be paired with mHealth applications when using them to conduct pain and spine intervention research, with implications for the clinical use of mHealth applications to monitor patient outcomes. By assessing recruitment success and survey completion rates, we provide a foundation for future researchers to optimize the design and use of mHealth applications in this field. Conclusions Varied recruitment methods paired with the use of a mobile application to track post-spine injection pain outcomes yielded medium to high overall enrollment and retention rates. In-person recruitment was more effective than remote recruitment for enrollment, and clinician recruitment may be more effective than non-clinician recruitment for post-enrollment survey completion rates. Future recruitment strategies should prioritize in-person interactions, and perhaps specifically in-person physician interactions, over remote research assistant approaches. Additional studies could explore barriers to retention in remote recruitment and examine alternative methods to improve long-term participation. Declaration of competing interest The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: Dr. Patricia Zheng reports financial support was provided by Nikolai Bogduk Young Investigator Grant. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Footnotes Research reported in this publication was supported by the Nikolai Bogduk Young Investigator Grant from the International Pain and Spine Intervention Society. Data will be made available upon reasonable request. References