Abstract
Introduction:
Participation in a clinical trial has been shown to influence medical adherence. We compared postoperative imaging completion and office visit attendance between patients enrolled in the Pediatric KIDney Stone (PKIDS) Trial and those who were not at two participating institutions. We hypothesized that study enrollment would improve adherence to postoperative follow-up.
Methods:
A retrospective review of patients (8–21 years) undergoing kidney stone surgery during the PKIDS study period was performed at two geographically different PKIDS institutions. Medical records were reviewed for completed imaging and follow-up office visits. The primary outcome was postoperative imaging acquisition within 8 weeks of surgery in alignment with the PKIDS protocol. Secondary outcomes included imaging acquisition by 16 weeks and postoperative office visit attendance by 8 and 16 weeks. Multivariable logistic regression assessed factors influencing adherence.
Results:
A total of 181 patients were included: 120 PKIDS vs 61 non-PKIDS. 51.9% of patients completed imaging at 8 weeks. PKIDS enrollment was not associated with imaging acquisition at 8 weeks or any secondary outcomes on univariate or multivariable analyses. A higher national Area Deprivation Index percentile (i.e., higher neighborhood-level deprivation) was associated with lower imaging acquisition on multivariable analysis (Odds ratio [OR] 0.98, 95% confidence interval [CI]: 0.97–0.99, p = 0.03). Conversely, private insurance was associated with higher attainment of imaging (OR 2.36, 95% CI: 1.09–5.12, p = 0.03) and office visit attendance within 16 weeks (OR: 3.08, 95% CI: 1.42–6.69, p < 0.01). There were no variables associated with 8-week postoperative office visits.
Conclusions:
PKIDS trial enrollment was not associated with increased adherence to postoperative follow-up in two geographically different PKIDS sites. This study supports the generalizability of the PKIDS study to non-PKIDS participants and future use of the PKIDS dataset to examine determinants of postoperative care in pediatric kidney stone disease.
Introduction
Kidney stone surgery follow-up is necessary to evaluate for stone clearance, assess for postoperative complications such as silent obstruction, and initiate metabolic workup and counseling of preventative strategies.1–3 Most data identifying barriers to kidney stone follow-up in children are extrapolated from adult literature, with reports of children completing postoperative workup 58–82% of the time.4–12 To better our understanding of pediatric stone disease, the Pediatric KIDney Stone (PKIDS) Care Improvement Network Trial was initiated to study the comparative effectiveness of surgical intervention.13–14 Important to postoperative imaging capture, the PKIDS study was embedded in clinical care and utilized a postoperative imaging protocol at 6 ± 2 weeks. Therefore, the PKIDS trial offers an opportunity to evaluate predictors of postoperative follow-up adherence.
To determine the generalizability of the PKIDS data set in future evaluation of postoperative adherence, we investigated if there was an inherent difference between follow-up in those participating vs not participating in the trial. The trial offered monetary incentives and had research coordinator support that might alter the natural course of follow-up. As two institutions participating in the PKIDS trial, we compared follow-up at our institutions between patients enrolled in PKIDS and those not to determine the impact of study participation on follow-up. Our primary outcome was adherence to postoperative kidney imaging within 8 weeks of surgery, as part of the PKIDS protocol. Our secondary outcome was postoperative kidney imaging acquired within 16 weeks of surgery, and attendance of postoperative follow-up appointment within 8 and 16 weeks. We hypothesized that participation in PKIDS would lead to better adherence to postoperative follow-up.
Methods
Study design and population
This bi-institutional, retrospective study was conducted at Nationwide Children’s Hospital (NCH, IRB# STUDY00003967) and the University of Alabama at Birmingham (UAB, IRB# 300012330). Patients between the ages of 8 and 21 years (age inclusion criterion for PKIDS trial) who underwent kidney stone surgery from May 1, 2020, to July 31, 2023, during each institution’s participation in the PKIDS trial were included. Patients were divided into (1) those who participated in the PKIDS trial (PKIDS cohort) and (2) those who did not (non-PKIDS cohort).
PKIDS participants were identified through prospective enrollment, whereas non-PKIDS patients were identified retrospectively from operating room schedules during the PKIDS enrollment timeframe. The non-PKIDS cohort included patients who either declined PKIDS participation, were missed during enrollment, or did not meet study inclusion criteria. Patients were excluded from PKIDS (1) if obtaining informed consent for enrollment would delay emergent care (such as obstructing stone with fever), (2) if they were non-Spanish or English speaking, or (3) if they were scheduled for simultaneous dual-modality stone treatments (i.e., shockwave lithotripsy with ureteroscopy). 13 Non-PKIDS participants followed each institution’s post-surgery follow-up protocol. Medical records were reviewed retrospectively to capture all relevant variables.
Outcome
The primary outcome was acquisition of postoperative imaging (renal ultrasound or non-contrast computed tomography [NCCT]) within 8 weeks of surgical stone removal, per PKIDS protocol. Both ultrasound and NCCT were acceptable modalities per institutional and PKIDS protocol standards; however, detailed modality documentation was variable across sites and, therefore, not included in the comparative analysis. Secondary outcomes included acquisition of postoperative imaging at 16 weeks, per the extended criteria of the PKIDS protocol, and attendance at follow-up office visits with a kidney stone provider (urologist or nephrologist) at 8 and 16 weeks.
Exposure
The main exposure variable was participation in the PKIDS trial. 13 The PKIDS cohort followed a structured follow-up imaging protocol as part of the trial, requiring kidney imaging at 6 weeks (+/− 2 weeks) after surgery, with follow-up extended to 16 weeks to improve capture of postoperative outcomes. Patients were compensated for completing patient-reported outcomes questionnaires, but not specifically for adherence with imaging acquisition. A dedicated research coordinator at each institution tracked protocol adherence and liaised with patients.
The non-PKIDS cohort followed each institution’s postoperative care standards; for NCH, imaging was scheduled and communicated to the family by the scheduling staff within 1–2 months after surgery, whereas the responsibility for follow-up scheduling rested with the family for COA. Follow-up office visits were not recorded as part of the PKIDS protocol, and thus postoperative office visits for the PKIDS cohort also followed institutional postoperative care standards.
Covariates
Covariates in the analysis included demographic, socioeconomic, and clinical factors that might influence adherence to postoperative imaging. These were: age, gender, insurance status (private, public, or self-pay), single-caregiver household status, initial presentation to the emergency department, prior surgical stone removal, National Area of Deprivation Index (ADI) percentile, and type of surgical intervention [ureteroscopy vs other (percutaneous nephrolithotomy, shock wave lithotripsy)]. We additionally collected data on each patient’s personal history of kidney stones, prior stone surgeries, family history of stones, reason for presentation to the emergency department (ED), diagnosis of septic stone at time of presentation, and need for ED visit postoperatively.
Single-caregiver household was identified by review of medical records closest to the time of surgery, with single-caregiver identified if: (1) directly stated the patient lived with a single parent; (2) one parent was listed as deceased with no other documentation of a stepparent; (3) parents were listed as divorced with no other documentation of a stepparent. Patients were classified as non-single (dual) parent households if there was specific hospital/clinic note documentation or if both parents were listed as contacts with matching addresses. Those lacking information on household status meeting the above requirements were excluded from analysis.
Distance traveled to the hospital was calculated from the listed home address to the location of the scheduled follow-up office visit. Both NCH and UAB have satellite office locations, and these locations were used as the hospital destination address when follow-up was scheduled at these locations. Google Maps was used to compute the distance traveled by road, which was recorded as “miles traveled.” Patients scheduled for telehealth follow-up were excluded from the distance analysis.
The ADI is a validated composite measure of neighborhood-level socioeconomic disadvantage, comprised of 17 U.S. Census variables such as income, employment, education, and housing characteristics. Each patient’s ADI national percentile was calculated by inputting their home address into the Neighborhood Atlas®. A low ADI represents less disadvantaged neighborhoods, whereas 100 equates to the highest level of disadvantage.15,16 ADI reflects broader community-level socioeconomic disadvantage, whereas private insurance represents individual-level access to healthcare, allowing both variables to capture distinct aspects of socioeconomic status.
Statistical analysis
Pearson’s chi-squared test was used to compare categorical variables between the cohorts. Continuous variables that were normally distributed (e.g., age, distance to hospital) were analyzed using independent samples t-tests, and non-normally distributed variables (e.g., ADI percentile) were compared using Mann–Whitney U tests. Binary logistic regression was performed to evaluate the association between predictor variables and adherence to postoperative imaging at 8 and 16 weeks, at both univariate and multivariable levels. A similar multivariable regression model was generated to determine adherence to attendance at follow-up office visits at 8 and 16 weeks.
The predictor variables included in the univariate models were: PKIDS enrollment, age, gender, insurance status (private, public, or self-pay), initial presentation to the emergency department, single-caregiver household status, prior surgical stone removal, National ADI percentile, and type of surgical intervention (ureteroscopy vs other types). The multivariable regression model included the same variables, regardless of statistical significance in the univariate analysis. Since the correlation between private insurance and National ADI percentile was moderate (r = 0.313, R2 = 0.098), both variables were retained in the multivariable analysis.
Multicollinearity among predictors was assessed using the Variance Inflation Factor, with a threshold of 10 indicating significant collinearity. None of the variables exceeded this threshold. The regression model achieved a p-value > 0.05 in the Hosmer–Lemeshow Test, indicating good model fit. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for both univariate and multivariable models to determine the effect of each variable on adherence. All statistical analyses were conducted using IBM SPSS Statistics 29.0 (IBM Corporation, Armonk, NY), with statistical significance set at p < 0.05.
Results
A total of 181 patients underwent kidney stone surgery: 120 in the PKIDS cohort and 61 in the non-PKIDS cohort. Patient demographics are listed in Table 1. The two cohorts were largely similar in demographics, apart from a higher proportion of ureteroscopy in the non-PKIDS group (96.7% vs 82.5%, p < 0.01). Overall, 51.9% of patients completed postoperative imaging within 8 weeks of surgery, and 67.4% attended an office visit within 16 weeks. Imaging acquisition increased to 69.6% by 16 weeks across both cohorts.
Comparative Overview of the Pkids and Non-Pkids Cohorts
ADI = Area Deprivation Index; ED = emergency department; IQR = interquartile range; n = number of patients; PCNL = percutaneous nephrolithotomy; PKIDS = Pediatric KIDney Stone (PKIDS) Care Improvement Network trial; SD = standard deviation; SWL = shock wave lithotripsy.
There was no significant difference between our cohorts for our primary outcome of imaging completion at 8 weeks (p = 0.40). Similarly, our two cohorts did not significantly differ in our secondary outcomes of imaging completion at 16 weeks and office-visit follow-up at 8 and 16 weeks (Table 2). The percentages of patients who attended both their visit and obtained imaging at 8 and 16 weeks were 39.8% (n = 72) and 62.4% (n = 113), respectively. Variables associated with obtaining imaging at 8 weeks are shown in Table 3. PKIDS enrollment was not significantly associated with imaging acquisition on either univariate or multivariable analysis. Private insurance was associated with a higher completion of imaging, whereas higher National ADI percentile (i.e., greater disadvantaged neighborhoods) and single caregiver households were significantly associated with lower completion of imaging on univariate analysis. A total of 173 (94.0%) patients with complete data on all variables were included in the multivariable regression analysis. On multivariable analysis, only a higher National ADI percentile was significantly associated with lower imaging acquisition.
Comparison of Imaging and Office Visit Follow-Up Between Pkids and Non-Pkids Cohorts Following Surgical Kidney Stone Removal
Univariable and Multivariable Analysis of Primary Outcome of Obtaining Imaging within 8 Weeks
CI = confidence interval; OR = odds ratio.
Multivariable analysis of secondary outcomes (Table 4) found that private insurance was significantly associated with both imaging acquisition (OR 2.36, 95% CI: 1.09–5.12, p = 0.03) and clinic follow-up attendance within 16 weeks (OR 3.08, 95% CI: 1.42–6.69, p < 0.01). Dual-parent household (p = 0.01), lower national ADI percentile (i.e., less neighborhood-level deprivation, p = 0.03), and private insurance (p < 0.01) showed a trend toward improved clinic attendance at 8 weeks on univariate analysis, but this was not significant in multivariable analysis. Similarly, higher National ADI (i.e., more disadvantaged neighborhoods) was significantly associated with lower imaging acquisition at 16 weeks (p < 0.01) and attendance of office visit at 16 weeks (p = 0.01) but was not significant on multivariable analysis.
Multivariate Analysis of Secondary Outcomes of Obtaining Imaging Within 16 Weeks and Office Visit Adherence at 8 and 16 Weeks Following Surgical Kidney Stone Removal
Asterisk (*) indicates significance on univariate analysis.
Discussion
Participation in a multi-institutional study did not alter postoperative follow-up (imaging acquisition or office visit) in patients with kidney stone disease at two institutions participating in the PKIDS trial. We hypothesized that trial-related financial incentives and administrative support not reflective of real-world practices would positively influence adherence to requested postoperative follow-up but were unable to demonstrate any significant differences in our study. When examining the influence of social determinants of health on postoperative adherence, we noted that private insurance and lower ADI (i.e., less socioeconomic disadvantage) correlated with better adherence. Our study is one of the first to link both individual-level and neighborhood-level social determinants of health to a pediatric chronic condition such as kidney stone disease. These results suggest a need to target better access to care for patients without private insurance living in areas of high socioeconomic disadvantage.
Our study serves as an a posteriori study of the generalizability of the PKIDS trial in evaluating follow-up adherence to the greater population of pediatric stone-forming patients. 17 Although our study was comprised of two institutions in two geographically diverse areas of the United States, the PKIDS trial included 31 medical centers across 28 unique health systems in the United States and Canada, capturing data on over 1200 patient participants. 13 Thus, the PKIDS cohort would be ideal for evaluating the impact of social determinants of health on postoperative follow-up. 13 Interestingly, the trial experienced its own difficulties in obtaining imaging within the initial protocol timeframe of 6 ± 2 weeks, resulting in an expanded endpoint of 16 weeks that we used in this study.
Participation in clinical trials can lead to certain biases or deviations in care that might limit their broader applicability. Specifically, patients in clinical trials have demonstrated alterations in their behavior, a phenomenon called “white-coat compliance” or Hawthorne effect and has been partially attributed to an awareness of being monitored. 18 This can lead to improved adherence to medical interventions. In a clinical trial of patients with uncontrolled hypertension, medication adherence increased in both trial and non-trial related medication usage during study participation. 19 This would suggest that those enrolled in a clinical trial are more involved overall with their medical conditions and treatment, specifically around the time of trial participation. However, this effect is not sustained, confirmed in a systematic review looking at compliance in blood pressure monitoring that found decreasing adherence after 6 months of time. 20
Limitations
We acknowledge that the two cohorts being investigated in this study are inherently different with explicit bias related to trial participation. PKIDS enrollment may be selectively biased toward patients who are interested in health outcomes. 17 The sample size is small; however, the purpose of this study was as a pilot investigation to validate a larger study with the PKIDS trial data. Given the sample size, the study may have been underpowered to detect modest but clinically meaningful differences in follow-up adherence between cohorts, particularly at the 16-week time point. Additionally, the small sample size may have underestimated any bias relating to trial participation and adherence to the protocol. Our study includes care provided during the COVID pandemic; parents may have chosen to follow up with their local provider during this time period, which may explain our overall lower adherence compared to what was previously reported.1,2,21 By keeping the study periods the same between cohorts, however, we felt we were able to control for this variable between our cohorts.
There was no formalized follow-up protocol for the non-PKIDS patients compared to those in the PKIDS trial. For these patients, there were differences in how postoperative follow-up is scheduled between the institutions, with UAB asking families to call postoperatively to schedule and NCH scheduling appointments for patients through a scheduling system. These differences may create heterogeneity in our results but also reflect the lack of standardization in follow-up protocols across institutions, thereby strengthening our findings as more reflective of real-world care. More importantly, it suggests that the PKIDS follow-up data is representative of postoperative activity of pediatric patients with stone disease at large and can be used for future studies.
Data collection varied between the two cohorts, with the PKIDS group being primarily collected prospectively, whereas the non-PKIDs participants were identified retrospectively. Lastly, although the decision to define adherence to imaging and follow-up was at 8 and 16 weeks per PKIDS protocol, use of this timing as a metric of adherence is relatively arbitrary. It does not account for specific circumstances that might limit a patient’s ability to accomplish imaging or travel to the clinic within this time frame.
Conclusion
Participation in the PKIDS trial did not impact postoperative imaging or follow-up office visit adherence in two institutions included in the trial. We validate that the larger PKIDS cohort can be used for the evaluation of predictors of kidney stone follow-up. Overall postoperative follow-up was low, necessitating further studies to identify predictors that might impact access to care and prioritize resources to these individuals.
Authors’ Contributions
C.M.C.T. and C.B.C.: Research conception and design. C.M.C.T., B.T.B., and C.B.C.: Data acquisition. B.T.B.: Statistical analysis. C.M.C.T., B.T.B., and C.B.C.: Data analysis and interpretation. C.M.C.T., B.T.B., and C.B.C.: Drafting of the article. C.B.C., B.T.B., J.S.E., B.A., and G.T.: Critical revision of the article. C.M.C.T. and C.B.C.: Supervision. C.B.C., B.T.B., J.S.E., B.A., and G.T.: Approval of the final article.
Footnotes
Funding Information
This work was supported in part through a Patient-Centered Outcomes Research Institute (PCORI) Program Award (CER-2018C3-14778). All statements in this report, including its findings and conclusions, are solely those of the authors and do not necessarily represent the views of the PCORI, its Board of Governors or Methodology Committee.
Submission Declaration and Verification
This work has not been published previously and is not under consideration for publication elsewhere.
Authors Disclosure Statement
The Authors declare that they have no affiliations with or involvement in any organization or entity with any financial interest in the subject matter or materials discussed.