Abstract
Introduction
In the past decade, technologies enabling direct patient care across distance have proliferated. A broad range of clinically important observations can now be acquired at locations convenient for patients and delivered to providers located almost anywhere. Among the various terms presently used to encompass the disparate technologies and service models now actively marketed for remote, direct patient care, we prefer “connected care.”
Questions about appropriate scope for use of different connected care models has prompted thoughtful discussion and recommendations addressing standards of care, 1,2 but contentious debate continues. 3 –5 (As of July 14,2015, the article by Goodnough 5 had generated 230 comments, many of which questioned quality of care via videoconference.) Although capacity to acquire clinical information with today's diagnostic tools is extensive, clinicians recognize that the scope of information required for appropriate clinical decision making and effective care is heavily determined by the specific clinical problem at hand. Similarly, the ability to connect and exchange some types of clinically relevant observations does not ensure that the richness of information—both its breadth and quality—is adequate for delivery of effective and efficient care of a specific patient and problem. The model's capacity, determined both by the available technology and how well that technology is used, might not encompass the appropriate range.
The mission of the Health-e-Access Connected Care Model is “to enable healthcare for families, when and where they need it, by providers they know and trust.” During the period addressed in this article, Health-e-Access focused on care of children with acute illness in schools, development centers, and childcare programs and during after-hours periods at convenient neighborhood sites. Effectiveness and efficiency of this model have been established 6 –10 and have been reviewed elsewhere. 11,12 Care was generally provided by a clinician from the child's primary care medical home. 13
Our goals herein are to promote replication of the Health-e-Access Connected Care Model by (1) identifying resources comprising this model, (2) describing service provided, (3) classifying diagnoses for completed visits on resources generally essential for appropriate care, and (4) assessing validity of this classification.
Materials and Methods
Observations began with implementation of telemedicine in childcare in May 2001 and continued with expansion of the model to elementary schools, a child development center, neighborhood community centers, and secondary schools (Table 1). Level of activity at these sites and collaboration of provider organizations changed over time. Observations ended June 2013. Clinical information was extracted, with identifying information removed, from telemedicine electronic medical records as in earlier research. 13
Visits by Type of Access Site by Year: May 1, 2001–June 30, 2013
Bold numbers indicate the year that service was initiated at a particular type of site. Percentages given are row percentages.
Included city school district, charter, and parochial schools.
Last 8 months of 2001 only.
First 6 months of 2013 only.
The process leading to a telemedicine visit was initiated when a problem was identified by the parents or the child site, and the program's coordinator engaged personnel, eliciting commitment from an appropriate provider to conduct the visit and coordinating communications among child site staff, parents, providers, and telemedicine assistants. Telemedicine assistants are trained personnel, generally with prior experience as nursing assistants or medical technicians, who collected clinical information at child sites about current concerns and past medical history. Collection of both history and digital media was guided by templates in telemedicine software. Templates were completed based both on conversations with parents and child site staff and on use of telemedicine devices capturing still images, video clips, and audio files. An all-purpose camera captured ear canal, tympanic membrane, nose, throat, eye, and skin images. The care model also included the capacity for collection and processing, when requested by a provider, of rapid streptococcal antigen tests, throat cultures, and skin or scalp cultures for fungus or bacteria.
Information collected by the telemedicine assistant was saved to a secure server where it was then accessible to the remotely located provider via personal computer. Information acquired via store-and-forward exchange was supplemented as required by the clinician, with both real-time video–audio interaction and telephone interaction with the child, child site staff, and parent. Real-time interactions enabled providers to observe the child's behavior, obtain additional history, explain assessment of the child's problem(s), discuss management recommendations, and have an opportunity, through this interaction, to establish trust. Video–audio interactions with parents occurred when a parent was present at the child site. When a parent was not at the child site, interactions between the parent and the provider occurred via telephone. (Although video–audio interactions with parents required parental presence at the child site during the study period, with the current availability of secure videoconference applications that operate on a range of smartphone platforms, this constraint no longer exists.) Interactions between the provider and the parent and between the provider and the child site generally occurred within 1 h of the time that concerns were brought to the attention of Health-e-Access staff. Clinicians faxed prescriptions to a pharmacy that in many instances delivered recommended medications and/or supplies to the child site. Further details on the Health-e-Access model are published elsewhere. 11
Measurement
We considered visits “completed” if diagnosis and management decisions were made and treatment was implemented to the same extent as they would be in a primary care physician's office. Completed visits did not require in-person physical examination, imaging, or laboratory testing beyond that incorporated in the model. Visits resulting in referrals to the emergency department were considered “noncompleted” because they reflected lack of capacity in the care model to acquire essential information or respond to it in timely fashion. Visits “completed with subspecialty referral” represented a subgroup of completed visits for which the provider decided that outpatient subspecialty management (e.g., ophthalmology, dentistry) was required, as they would in a primary care office setting.
Primary diagnoses were classified on “key resource requirements,” defined as resources usually needed to collect information to establish the primary diagnosis and to manage the problem regardless of setting (in-person or telemedicine). We envision that this classification might facilitate implementation and dissemination of this care model by demonstrating the importance of various Health-e-Access components and by providing payers with an evidence base that helps in determining whether a connected care model is being used appropriately. A primary diagnosis of acute otitis media, for example, requires a model including an electronic otoscope. This diagnosis cannot be used to justify payment if the model used included only texting or only video–audio acquisition and exchange of information. The two clinicians among authors (K.M.M. and S.R.) performed this classification. Both were experienced in caring for patients using Health-e-Access.
Actual management decisions made by participating clinicians were used to validate these classifications. If the visit was completed, we assumed that resources available through Health-e-Access were adequate for delivering high-quality care. On this basis, when a visit with a diagnosis that investigators classified as often beyond the scope of Health-e-Access (i.e., resources often required were not part of the model) was, nevertheless, completed, the classification was considered to be falsely negative. As an example, consider a visit assigned a primary diagnosis of gastroenteritis. Urinalysis, which helps evaluate hydration status, was not available through Health-e-Access. For a completed gastroenteritis visit, this classification was considered a false-negative based on the assumption that the clinician was reassured by history (e.g., normal fluid intake and output) or visual observation (e.g., normal activity level) that dehydration was not a problem.
To enable comparison of service delivered through this care model with service delivered in office and in emergency department settings throughout the United States, we tabulated primary diagnosis based on ICD-9 codes selected by the provider.
Results
Between May 1, 2001 and June 30, 2013, in total, 13,812 telemedicine visits were initiated. Among these, 13,560 (98.2%) were completed, and 252 were not. Over the 48 quarters in the 12 years of observation, the proportion of visits completed ranged from a low of 93.9% to a high of 100%. Mean (standard deviation) age of children at completed visits was 5.3 (4.0) years. Children 12 months of age or less represented 9.5% of all visits; those 24 months or less, 25%; and those 60 months or less, 58.1%. The remaining 41.9% of visits were for individuals beyond their fifth birthday and under 22 years of age.
For the 252 visits not completed, providers explained failure to complete as follows: diagnosis or management beyond the scope of the model (e.g., appeared too ill), 27.0%; unwilling to complete for administrative reason (e.g., child was overdue for well childcare), 17.1%; poor quality of images or lung sounds, 13.1%; child not sufficiently cooperative, 5.2%; and language barrier, 0.4%. For the remaining 37.3% of noncompleted visits, providers failed to provide an explanation. Among the 27.0% considered by providers to be beyond scope, most (31 of 68) were not assigned a diagnosis. Among the 37 assigned a diagnosis, 21 (56.8%) were wheezing.
Tables 1 –3 display the distributions of visits by site over the 12 years of observation, primary diagnoses for completed encounters, and principal resources required to manage these diagnoses, respectively. Principal resources required most commonly were those for examination of the ear canal and tympanic membrane, accounting for 31.8% of completed visits. These resources included tools and personnel for cerumen removal as well as an otoscope for capturing tympanic membrane images. Primary diagnoses associated with this requirement were acute otitis media, cerumen impaction, chronic suppurative otitis media, ear canal foreign body, ear drum perforation, ear pain, eustachian tube dysfunction, hearing deficit, otitis externa, otitis media with effusion, and tinnitus.
Distribution of Primary Diagnosis for 13,560 Completed Visits
Includes all diagnoses comprising 0.9% of the total or greater.
Distribution of Telemedicine Visits by Key Resource Requirement
Often reqiring cerumen removal.
Excluding retinal exam.
Auscultation of lungs.
Ten diagnoses accounted for 75.8% of these 306.
May require the clinician to palpate, manipulate, or perform a neurologic exam.
Such as evaluation for child abuse or neglect.
Three hundred six visits (2.3% of all 13,560 completed) were completed despite a primary diagnosis that had been classified as at least sometimes beyond the model's scope. Resources sometimes needed, but not part of the model, were hands-on exam by a clinician, behavioral evaluation, technology not included in the model, subspecialist involvement, and specialized history (Table 3).
Discussion
Completed Visits and Effectiveness
The range of sites used suggests the breadth of potential applications. The high completion rate demonstrates effectiveness; telemedicine visits that do not yield diagnosis and management decisions and result in treatment, when indicated, are not as useful as most in-person visits. It should be understood, however, that telemedicine encounters followed by in-person visits may still be useful to patients if between-site collaboration is sufficient to expedite in-person care. In such instances the telemedicine visit becomes, effectively, a second-level triage. Thus, as long as initial triage directs true emergencies to the emergency department, an obligation that can be met by following long-established telephone triage guidelines, 14 all other problems that are generally seen in person can be safely directed to a site with Health-e-Access resources.
We believe that continuity within the medical home, previously established to be high for implementation of the Health-e-Access program, 13 contributes substantially to the high level of visit completion. Given prior parent–provider interactions, the likelihood of mutual trust and understanding increases, and the provider may feel a stronger commitment to provide convenient care. Moreover, the provider may be more likely to trust the parent to recognize and respond to subsequent symptoms of deterioration.
Primary Diagnosis
The distribution of primary diagnoses (Table 2) is similar to that for children seen in emergency departments 15 and, even more so, in primary care offices throughout the United States. 16 The 20 most common conditions are frequently associated with fever and change in behavior. This association holds for the top three diagnoses, as well as two more among the top six (streptococcal pharyngitis, nonspecific viral illness). Together, these five primary diagnoses account for 48.2% of those managed. Behavior changes may be subtle, especially in the first few years of life. 17 –20 Parent–provider interactions limited to verbal communication undoubtedly provide a less reliable basis for judgments about the possibility of severe bacterial infection or respiratory distress than in-person evaluation. Video–audio interactions seem much more likely to achieve the validity of in-person assessment in these important judgments than other forms of connected care, such as audio-only telephone. Evidence supports this view; observations of appearance, activity level, and responsiveness have been shown to be reliable and valuable in differentiating serious from self-limited conditions in young febrile children. 17 –20
Video–audio communication also seems more likely to enhance the ability of the provider to impart trust. This is especially important in situations where the parent and the provider have had little prior contact, where the provider has nothing to offer to shorten the course of the illness, and where the service offered, fundamentally, is reassurance that the child's problem is self-limited. Finally, inclusion of video–audio communication may also be important to provider confidence and satisfaction, by enhancing the provider's ability to “read” a parent's understanding, acceptance, and appreciation. The importance of eye contact to understanding another's emotional state is well supported. 21,22 Accordingly, we believe a connected care model designed to address a comparably broad range of acute childhood problems requires capacity for video–audio interaction both for safety and to ensure effective communication.
Classification of Resource Requirements
Among the 13,560 visits completed, 2.3% were assigned a primary diagnosis classified as often requiring a resource not part of Health-e-Access. These were “false-negative” classifications; despite a primary diagnosis classified as often beyond the scope of Health-e-Access, these visits were completed successfully. The classification performed well in the study population; if regulatory agencies or payers required further justification (beyond diagnosis itself) for services based on visits flagged in a similar population as possibly inappropriate, this would be required for only about 2.3% of completed visits.
Visits not completed visits despite diagnoses classified as almost always manageable represented “false-positive” classifications. Because many illness episodes with visits not completed were not evaluated to the extent that a diagnosis was assigned, determination of the false-positive rate based on classification of diagnoses is not possible. We can, however, establish an upper bound for false-positives based on visits initiated but not completed. Taking the worst-case assumption that all of the 252 noncompleted visits had a diagnosis classified as appropriate for the Health-e-Access model, false-positives would constitute 1.8% (252 of 13,812) at most. We know that 1.8% is a substantial overestimate of such errors because we found that a large proportion of reasons for noncompletion had nothing to do with fit of the model (e.g., provider unwilling to complete for administrative reasons). Also, we know from prior research that such errors would rarely result in adverse events. 10 Almost always, payment for a visit completed using this model would be payment for a service successfully delivered.
Limitations
Resource requirements (Table 3) and the distribution of primary diagnoses (Table 2) validate the decisions to include peripheral devices that enable (1) diagnostic-quality images of the tympanic membrane, ear canal, throat, eye, skin, and scalp, (2) diagnostic-quality lung sounds, and (3) capacity for point-of-care rapid streptococcal antigen testing and collection of specimens for throat and fungal cultures. In focusing on the type of resources required to manage the primary diagnosis, however, this classification ignored the importance of model components needed to manage other problems possibly present or to rule out other conditions that might manifest signs and symptoms similar to the primary problem.
Consider the common problem of an infant with fever. Evidence suggests that video–audio observation of a febrile infant enables observations of subtle behaviors that distinguish serious bacterial illness from common, self-limited viral illness. 23 A claim that connected care depending on telephone or text messaging would be reliable in this situation lacks face validity. Thus, understanding the design and implementation of a connected care model is essential to understanding its capacity (i.e., the range of clinical problems addressed appropriately).
Furthermore, consider the primary diagnosis, upper respiratory tract infection. Required resource category in this instance was “other upper respiratory.” Among many children with manifestations of upper respiratory infection, ear examination may be important to rule out acute otitis media. Also, if the child has a troublesome cough or tachypnea, lung auscultation should be performed to address the possibility of conditions such as bronchiolitis, asthma exacerbation, or pneumonia.
Nevertheless, in addressing resource requirements for over 13,500 visits, it is unlikely that those conditions sometimes appearing as second or third diagnoses were not encountered as a primary diagnosis. For both these reasons, the distribution of resource requirements does not reflect the relative importance of model components, nor can it be used to justify exclusion of Health-e-Access components in alternative models designed to address a narrower range of clinical problems.
We do not know the “universe” of illness episodes that arose among children in participating sites. Observations reported herein did not capture situations where child site staff or parents, for any reason, decided not to seek care through Health-e-Access.
Conclusions
Defining the scope of a particular connected care model is essential for clinical, regulatory, and financing purposes. Appropriate scope is determined by the capacity, of resources included in that mode, for information acquisition and exchange. Reimbursement based on the resource requirement classification of diagnoses discussed herein would support payment for a large majority of illness episodes presenting in pediatric primary care settings. Limiting reimbursement to these diagnoses would limit the value of this model relatively little.
Regulations and quality metrics for connected health should be developed in light of fundamental attributes of the medical profession. Professions have evolved as society's response to the reality of specialized knowledge, which members of the general public cannot be expected to master. 24 Components of a connected care model and broad parameters for scope of practice can be specified. Beyond that, however, application of a connected care model to specific illness episodes should be left to the judgment of a medical professional, committed to serve the best interest of the patient, as it has been since invention of the telephone enabled the first connected care model over a century ago. 25 –27
Replicating Health-e-Access with a service offering similar effectiveness and safety would benefit families and communities substantially. Resources used in Health-e-Access may be used to guide both selection of technology and configuration of organizational architecture for replication. Replication requires an information-rich, connected care model with comparable components and financing models that incentivize and regulations that enable its implementation.
Footnotes
Acknowledgments
Sources of support for the research that is referenced in this article and that contributed to development of opinions expressed herein are as follows: Agency for Healthcare Research and Quality (grants R18 HS018912, R01 HS016871, and R01 HS15165); New York State Healthcare Foundation (Optimizing Access to Avoid Emergency Department Use for Non-Emergency Problems); New York State Health Department (Telemedicine to Optimize Access to Primary Care); Maternal and Child Health Bureau, Health Resources and Services Administration (grant R40 MC03605); the Robert Wood Johnson Foundation (RWJF Local Initiative Funding Partners); and U.S. Department of Commerce Technology Opportunities Program (Telemedicine in Daycare). RWJF funds were matched by funds from several anonymous donors and from 11 Rochester-area foundations, including the Rochester Area Community Foundation, Daisy Marquis Jones Foundation, United Way of Rochester and Monroe County, Halcyon Hill Foundation, Rochester's Child, Gannett Foundation, Marie C. and Joseph C. Wilson Foundation, Fred and Floy Wilmott Foundation, Weyerhaeuser Company Foundation, Feinbloom Family Supporting Foundation, and Frontier Telecommunications Corporation.
Disclosure Statement
At one time, K.M.M. and N.E.W. held financial interests in TeleAtrics, Inc., which purchased the software used to deliver services described in this article from Trifecta Technologies of Allentown, PA. TeleAtrics software was then acquired by Trifecta Technologies in August 2008. K.M.M. and N.E.W. were co-founders of TeleAtrics Inc. and, as a result of the August 2008 sale, were to receive a small percentage of potential revenues that might accrue to Trifecta through July 2013 from this product. Neither K.M.M. or N.E.W. received any funds from Trifecta following the time of the sale of TeleAtrics in August 2008. C.A. and S.R. declare no competing financial interests exist.