Abstract
Introduction
The COVID-19 pandemic greatly expanded the use of telehealth in nephrology, and telehealth remains a popular modality for kidney care. However, there are no consensus guidelines on the implementation of telehealth in ambulatory nephrology care. To examine this field further, we examined the association between clinic encounter modality (in-person versus telehealth) with provider actions to advance patient care plans in one academic nephrology practice.
Methods
Electronic health records from 26,170 telehealth video and 11,558 in-person encounters from January 2020 to December 2024 were analyzed. Mixed effects models were used to assess the association between encounter modality and medication prescription change rates, order rates, and patient instruction documentation frequency.
Results
Compared to in-person encounters, telehealth video visits served patients of younger age (60.4 years (SD = 18.1) vs. 66.7 (17.4), p < 0.001) and with less severe kidney disease (estimated glomerular filtration rate 51.9 ml/min/1.73 m2 (SD = 30.2) vs. 46.1 ml/min/1.73 m2 (26.6), p < 0.001). Accounting for encounter, patient, and provider characteristics, telehealth modality was independently associated with 23% fewer medication prescription changes (incidence rate ratio (IRR) = 0.77, 95% CI 0.74–0.80), 38% fewer orders (IRR = 0.62, 95% CI 0.57–0.68), and 84% lower odds of patient instructions documentation (odds ratio = 0.16, 95% CI 0.09–0.31) compared with in-person encounters.
Conclusion
These data suggest that providers use telehealth video and in-person encounters to advance patient care plans differently. Acknowledging these differences will be crucial to establish best practices for telehealth utilization in ambulatory nephrology care and assure ongoing high-quality care delivery.
Introduction
Since the COVID-19 pandemic, telehealth, broadly defined as a healthcare provider's use of information and communication technology in the delivery of clinical services, 1 has emerged as a valuable and enduring modality for delivering ambulatory kidney care. Its appeal is multidimensional. Economically, telehealth is more cost- and time-efficient than in-person visits as it reduces the need for patient and provider travel, reduces clinic overhead, and improves patient flow.2–5 Beyond logistical considerations, by requiring patients to conduct activities at home that normally occur during an office visit (e.g. ambulatory blood pressure and glucose monitoring) and reporting to their clinicians, virtual care delivery can promote healthy behaviors and support patient autonomy by encouraging self-monitoring, which are known to benefit individuals with chronic kidney disease (CKD).6,7 Nephrology is especially well-suited for telehealth. Kidney care relies heavily on objective, remotely accessible data such as blood pressure, estimated glomerular filtration rate (eGFR), electrolytes, acid-base status, and proteinuria for disease evaluation and clinical decision-making. Telehealth also has the potential to alleviate the growing supply-demand mismatch between rising kidney disease prevalence and a stagnant nephrology workforce by increasing efficiency in healthcare delivery.8,9 Additionally, telehealth can improve equitable access to nephrology care. Patients who live in rural areas have been shown to experience worse kidney outcomes than their urban counterparts due, in part, to a lack of available nephrologists and subspecialty infrastructure9–11; telehealth may help bridge this care gap. For these many reasons, nephrologists are among the highest utilizers of telehealth, and physician satisfaction with telehealth has generally been favorable.12–15
Telehealth encompasses multiple modalities, including synchronous telehealth video encounters and audio-only telephone visits, which differ in technological requirements, accessibility, and clinical capabilities. Telehealth video encounters allow for greater interpersonal connection and some limited visual examination but have higher technological requirements than audio-only encounters. In direct comparisons, telehealth video visits are preferred by both patients and providers, and have been associated with superior clinical and operational outcomes compared to audio-only encounters.16,17 However, there are currently no consensus guidelines for the clinical implementation of telehealth video technology in ambulatory nephrology care delivery, despite widespread use. 18 To explore this field further, we examined how one academic nephrology practice uses telehealth video visits and compared the association between ambulatory nephrology clinic modality (in-person versus telehealth video), with provider actions that advance patient care plans, including medication prescription changes, orders placed, and patient instructions documentation.
Methods
Study population and measures
A total of 57,372 encounters from 1 January 2020 to 31 December 2024, at the adult general nephrology clinic of one academic health center in California were analyzed. This date range was chosen as, prior to this period, telehealth video was not a frequently utilized modality at this institution. In response to the COVID-19 pandemic and nationwide telehealth policy changes, administrative guidance was issued at this center on 28 February 2020 recommending utilization of telehealth “whenever possible.” Concurrent infrastructure and workflow changes substantially expanded telehealth video access and established it as a primary rather than supplementary mode of care delivery. These shifts were reflected in utilization patterns: telehealth video comprised only 6.1% of encounters in 2019, rising to 66.8% in 2020, and it has remained a key ambulatory care modality since then (Figure 1). 2019 data were extracted from the electronic health record and reviewed but were excluded from the final analysis due to the markedly lower volume and fundamentally different nature of telehealth utilization during that period.
Telehealth use over time. Percentages represent yearly average proportion of telehealth use.
Next, canceled and “no-show” appointments (n = 17,879) were excluded from the analysis. Audio-only telehealth encounters (n = 1178) were also excluded from this analysis, as they lacked visual context for providers and consistency in setting. Atypical encounters (e.g. solely for ambulatory blood pressure monitoring, peritoneal dialysis follow-up, etc.; n = 587) were also excluded from this analysis as they were determined to be qualitatively different from standard ambulatory nephrology clinic encounters. A total of 37,728 encounters were used for cross-sectional analysis.
Encounter date, provider of record, encounter modality (in-person vs. telehealth video), encounter type (new patient or follow-up), and billing documentation were extracted from the medical record. Patient demographics (age, sex, race/ethnicity, insurance), medical history (hypertension, diabetes, coronary artery disease, congestive heart failure, cerebral vascular accidents), patient portal enrollment, and medication lists were evaluated at each encounter. Current Procedural Terminology 19 billing codes were also extracted from encounters; codes 99214 and 92204 described clinic encounters of moderate complexity while codes 92215 and 92205 characterized high complexity visits. Lab data were attributed to the nearest nephrology encounter using a 1-year lookback period. The 2021 CKD-EPI equation was used to calculate eGFR from extracted serum creatinine values. 20 A 6-month lookback period was used for patient height and weight measurements.
The primary predictor for this analysis was encounter modality, either telehealth video or in-person. Telehealth video visits were conducted through Zoom Communications technology (San Jose, USA). The outcomes assessed were encounter-level provider actions, defined as medication prescription changes, order placement, and patient instructions documentation. Medication prescription changes were defined as the number of new prescriptions, prescription cancelations, or dose changes associated with an encounter. Reordering medications at the same dose was not counted as a medication prescription change. Similarly, changes to historical or patient-reported medications, such as over-the-counter medications or medications prescribed by other institutions, were excluded. Provider orders placement was defined as the number of non-medication orders associated with an encounter. Laboratory orders were excluded from this analysis as they were often placed asynchronously from patient encounters. Similarly, indirect and non-patient effecting orders, such as nursing orders and provider notification orders, were excluded. Patient instructions documentation was defined as the presence or absence of a patient-facing note filed into the electronic health record and encounter after the visit summary. Patient instructions were not required to close or bill for encounters during this time. Whereas medication prescription changes and orders were analyzed as numerical counts, patient instructions documentation was analyzed as a binary.
Statistical analysis
Two sample t-tests, two proportion t-tests, and Chi-squared tests were used to compare baseline characteristics and outcomes between encounter modalities. Generalized linear mixed effects models using negative-binomial regression were used to assess the association between encounter modality and medication prescription changes and orders, accounting for patient- and provider-level random effects and adjusting for patient factors (age, sex, race, insurance, patient portal enrollment) and encounter factors hypothesized to be associated with both encounter modality and number of provider actions (eGFR, baseline number of medications at the beginning of each encounter prior to any medication prescription changes, encounter type (new patient or follow-up), history of diabetes, and history of hypertension). A similarly adjusted mixed effects model using logistic regression assessed the association between encounter modality and patient instruction documentation; this model was further adjusted for the number of medication changes and orders placed during each encounter. These mixed effects models were used to calculate the relative number of medication changes, orders, and patient instructions among telehealth and in-person encounters. The incidence rate ratio (IRR) was calculated for numeric outcomes of medication changes and orders. The odds ratio (OR) was calculated for the binary outcome of patient instructions documentation. Statistical analysis and image generation were done with R 21 and Microsoft Excel 22 software.
Results
Patient and encounter characteristics
A total of 37,728 encounters, 11,558 in-person and 26,170 telehealth video, were included in this analysis describing 7262 unique adult patients who were seen an average of 5.2 times by 26 providers from 2020–2024. 1291 patients had exclusively in-person encounters, 3270 only participated in telehealth encounters, and 2701 patients experienced both in-person and telehealth encounters over the study period. Summary statistics describing the encounters by modality are shown in Table 1. Telehealth encounters, on average, served patients of younger age (60.4 (SD = 18.1) vs. 66.7 (17.4)), with higher eGFRs (51.9 ml/min/1.73 m2 (SD = 30.2) vs. 46.1 ml/min/1.73 m2 (26.6)). Patients with telehealth encounters also had lower rates of hypertension (88.3% vs. 92.2%), diabetes (40.5% vs. 52.3%), coronary artery disease (27.3% vs. 33.5%), and stroke (21.1% vs. 24.9%), and were taking fewer medications at the start of the visit (9.4 (SD = 8.2) vs. 9.8 (8.0)). Finally, telehealth encounters were less frequently new patient visits (11.2% vs. 17.9%) and were less frequently coded as high complexity (64.6% vs. 78.2%) compared to in-person visits (all p < 0.001; Table 1).
Encounter characteristics for telehealth video versus in-person encounters from 2020–2024.
eGFR: estimated glomerular filtration rate; CPT: Current Procedural Terminology.
BMIs > 100 and <10 were excluded. 3498 and 4583 encounters were missing eGFR and BMI data, respectively.
Number of provider actions: medication prescription changes, non-medication orders and patient instruction documentation
On average, 0.99 (SD = 1.6) medication prescription changes were made per encounter. More medication changes were made during in-person visits than telehealth visits (1.2 (SD = 1.6) vs. 0.92 (1.5), p < 0.001). The proportion of medication changes that were new medication prescriptions was also higher during in-person encounters compared to telehealth video encounters (35% (SD = 48%) vs. 29% (46%), p < 0.001, Figure 2).
Mean number of medication changes per encounter by encounter modality.
On average, 0.20 (SD = 0.40) non-medication orders were placed per encounter. More non-medication orders were placed during in-person encounters than telehealth visits (0.27 (SD = 0.44) vs. 0.17 (0.37), p < 0.001)). The majority of orders were either referrals (59.4%) or imaging studies (32.4%).
Patient instructions were more frequently documented during in-person encounters compared to telehealth encounters (93.2% vs. 78.6%, p < 0.001). When documented, there was no difference in length of patient instructions between telehealth and in-person encounters (612 characters (SD = 632) vs. 605 (619), respectively; p = 0.34).
Association of encounter modality with provider actions
In hierarchical negative-binomial and logistic regression models and accounting for clustering of encounters within patients and providers, telehealth visits were independently associated with 23% fewer medication prescription changes (IRR = 0.77, 95% CI 0.74–0.80), 38% fewer orders (IRR = 0.62, 0.57–0.68), and 84% lower odds of patient instruction documentation (OR = 0.16, 0.09–0.31) (Figure 3).
Forest plots of select variables used in mixed effects models for (A) medication changes, (B) orders, and (C) patient instructions outcomes. The full set of predictor variables used for mixed models is described in Methods. For the categorical variables of Telehealth, Male Sex, and First Visit, ratios are given in comparison to reference values of “In-person,” “Female Sex,” and “Follow-up visit,” respectively. Ratios for continuous variables are per SD of the respective variable. Error bars represent the 95% CI. Each outcome panel is plotted on an individual x-axis.
Apart from encounter modality, other factors were found to be associated with differential encounter-level provider actions. Notably, older patient age was associated with fewer medication changes (per SD; IRR = 0.90, 95% CI 0.87–0.93 and fewer orders (IRR = 0.93, 0.88–0.98), but wasn’t associated with differential patient instructions documentation rate (OR = 0.98, 0.91–1.05). In addition, less severe kidney disease, characterized by higher eGFR, was associated with fewer medication changes (per SD; IRR = 0.94, 95% CI 0.92–0.96) and fewer orders (IRR = 0.87, 0.84–0.91), but increased patient instructions documentation rate (OR = 1.09, 1.04–1.15). History of hypertension was associated with more medication changes (IRR = 1.51, 95% CI 1.41–1.62) and increased patient instructions documentation (OR = 1.19, 1.02–1.38) but was not associated with differences in order placement (IRR = 0.89, 0.80–1.00). History of diabetes was associated with more medication changes (IRR = 1.16, 95% CI 1.11–1.20) but no change in orders (IRR = 0.93, 0.86–1.00) or patient instructions documentation (0.94, 0.84–1.04). First visits were associated with higher rates of all outcomes (IRR = 1.06, 95% CI 1.01–1.12; 4.46, 4.10–4.86; OR = 1.18, 1.02–1.35 for medication changes, orders, and patient instructions documentation).
Discussion
In one academic center's nephrology practice, telehealth video modality was independently associated with fewer medication changes, fewer orders placed, and fewer patient instructions documentated compared to in-person nephrology care delivery. Our findings corroborate results of prior studies in primary care contexts in which telehealth encounters were also associated with fewer medication prescriptions and orders compared to in-person visits. In a national survey of telehealth in primary care settings during the COVID-19 pandemic, Alexander et al. noted fewer new medication prescriptions for cholesterol and blood pressure in telehealth encounters versus in-person encounters (38.8–39.3% vs. 44.9–47.1% of encounters with a new prescription during Q1-Q2 of 2020). 23 Wabe et al. performed a similar study in Australia and found a similar difference between medication prescription rates in telehealth vs. in-person encounters (33.0% vs. 39.3%, respectively). 24 Reed et al. evaluated a Californian cohort and found an 8.3% difference in medication prescriptions (38.4% vs. 46.8%) and an 8.5% gap in imaging orders (11.9% vs. 20.5%) between telehealth and in-person encounters. 25 Casey et al. report a similar trend in pediatric primary care, with a 10.3% gap in medication prescriptions (29.5% vs. 39.8%) and a 4.5% gap in imaging orders (4.0% vs. 8.5%). 26 Research on patient instructions documentation and studies regarding nephrology are sparse.
There may be inherent qualities about the telehealth encounter modality that lead to fewer provider actions. First, objective data like vital sign readings, physical exam signs, and point-of-care lab tests are generally less available in remote encounters. Providers may feel less confident in prescribing medications or placing orders without this information. In addition, telehealth encounters often take place outside of designated healthcare settings, so patients and providers may be less focused, diminishing clinician interest or motivation to make therapeutic changes or document patient instructions. Compared to in-person encounters, a higher proportion of patients who engage in telehealth video visits are likely to have smartphones, reliable internet access, and more formal education, 27 factors that may reassure providers that formal patient instructions are less necessary, and that impromptu follow-up communication between patient and provider could occur if required. The largest discrepancy observed between encounter modalities was in patient instructions documentation. Although speculative, this finding may be explained by differences in clinical workflow. During in-person visits, patient instructions are typically printed and reviewed with patients by clinic staff during the checkout process. Because this workflow does not occur during telehealth video encounters, providers may perceive less utility in documenting formal patient instructions. Finally, given the logistical challenges and costs associated with in-person encounters, and the precious nature of true face-to-face time, providers may be more compelled to perform additional encounter-level actions during in-person interactions than during telehealth encounters. However, the associations between encounter modality and fewer provider actions seen in this analysis may also be due to unmeasured confounding. This analysis was unable to account for contextual patient and provider attitudes and biases about virtual care delivery, which could have influenced study results.28,29 Patient comfort and familiarity with telehealth technology were not accounted for either, though models were adjusted for patient portal enrollment. In addition, some element of confounding by indication may be contributing to the study results. As seen in Table 1, telehealth video visits were generally used for less complex follow-up visits with healthier patients, in whom a lower number of medication changes, and orders and patient instructions would be expected. In other words, this discrepancy in provider interventions may reflect clinical appropriateness rather than inherent qualities of different encounter modalities. However, the associations between telehealth video encounter modality and fewer medication prescription changes and orders and fewer patient instructions were persistent throughout the study period, which included the COVID-19 pandemic (approximately 2020–2023), when many encounters were systematically encouraged to be converted to telehealth video, regardless of medical complexity or other contextual factors. This suggests that patient and physician discretion of telehealth video utilization may not be the sole explanatory factor for these results.
This study adds to prior literature on this subject by examining the use of telehealth video and in-person encounters in a specialty context, where patient variability is relatively narrower compared to primary care. Though kidney care spans a broad spectrum of conditions, most nephrology patients are seen for CKD management. By including eGFR in the analysis, the confounding factor of kidney disease severity was largely accounted for in this study, thus facilitating a more robust comparison of encounter modalities for ambulatory care delivery. This study also evaluated medication prescriptions as a continuous variable rather than a binary variable as was done in prior studies,23–26 enabling a more nuanced assessment of care interventions. Finally, this study assessed telehealth video's performance over a wide timeframe, including years 2023–2024, when pressures from the COVID-19 pandemic lessened, potentially offering a more generalizable understanding of telehealth's use for the delivery of ambulatory nephrology care.
This analysis has several weaknesses. First, this study was limited to a single academic center with a robust telehealth video infrastructure which limits generalizability. In health care centers where telehealth video is less feasible, there may be even more dramatic differences in baseline population characteristics between telehealth video utilizers and those who perform visits in-person, which limits direct comparisons. In addition, study outcomes were limited to structured elements linked to each encounter in the electronic health record. Relevant unstructured or unlinked actions were not captured. For example, this study could not detect if a provider delivered patient instructions through a patient portal message or other alternative means in lieu of formally documenting patient instructions in a note. Also, this study is subject to random measurement error in two of the outcomes. The action of discontinuing a medication and reordering it at a new dose was considered two medication changes, while the act of modifying a medication's dose was considered one medication change, though these actions are functionally equivalent. Similarly, despite having little clinical relevance, incomplete orders were included in this analysis because the reason for incompletion could not be consistently ascertained from chart abstraction. For example, because we could not differentiate between referrals canceled because of patient and provider decision-making during the encounter and those eventually administratively canceled due to never being scheduled, orders of all statuses were counted. However, it is unlikely that these measurement errors are differential across encounter modality. Additionally, there is a small proportion of medications and orders that were predominantly ordered during in-person contexts, notably vaccines and other clinic medications (i.e. erythropoietin-stimulating agents), which may have skewed results, though the absolute difference in number of these prescriptions per encounter was relatively small (0.0094 more new vaccine prescriptions per encounter and 0.014 more new erythropoeitin-stimulating agent prescriptions per encounter).
In summary, this analysis corroborates previous results and suggests that there are differences in provider actions between telehealth video and in-person encounter modalities, independent of measured differences between patients and encounters. Future studies investigating this difference will be critical to establish best practices for telehealth video utilization as an option for ambulatory nephrology care delivery.
Footnotes
Acknowledgements
The authors acknowledge Salman Rahman, MD, who contributed to this project as a data analyst.
Ethical considerations
This study was performed in accordance with the World Medical Association's Declaration of Helsinki. This human study was approved by the University of California San Francisco Institutional Review Board—approval number: 24-42349. All participants were adults. As determined by the University of California San Francisco Institutional Review Board, informed consent was not required because of the scope and minimal-risk nature of this study.
Consent to participate
Not applicable.
Consent for publication
Not applicable.
Author contributions
Research area and study design were done by KS, LP, AO, LL, and DT. Statistical analysis was done by KS, LP, and AO. Supervision or mentorship was done by AO, LL, and DT. Manuscript writing and review were done by KS, LP, AO, LL, and DT. Each author contributed important intellectual content during manuscript drafting or revision and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved. KS takes responsibility that this study has been reported honestly, accurately, and transparently; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned have been explained.
Funding
The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute of Diabetes and Digestive and Kidney Diseases, (grant number 5TL1DK139565-02).
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
The data generated during and/or analyzed during the current study are not publicly available due to the presence of protected health information (patient demographics, encounter data) but are available from the corresponding author on request.