Telemedicine Reduces Absence Resulting From Illness in Urban Child Care: Evaluation of an Innovation
http://www.100md.com
《小儿科》
Department of Pediatrics School of Nursing
Department of Biostatistics and Computational Biology, University of Rochester Medical Center, Rochester, New York
ABSTRACT
Background. Common acute illness challenges everyone involved in child care. Impoverished inner-city families, whose children are most burdened by morbidity and whose reliance on child care is most important, are those least equipped to deal with this challenge.
Objective. To assess the impact of telemedicine on absence from child care due to illness (ADI).
Design/Methods. A before-and-after design with historical and concurrent controls was used to study ADI in 5 inner-city child care centers in Rochester, New York, between January 1, 2001, and June 30, 2003. Enrollment averaged 138 children per center, of whom Medicaid covered 66%. Center 5 provided only concurrent controls. Telemedicine service began in the first 4 centers in a staggered fashion starting in May 2001. Baseline data on ADI before availability of telemedicine were collected in each center for a minimum of 18 weeks. The telemedicine model for diagnosis and treatment of common acute problems involved both real-time and store-and-forward information exchange between a child and telemedicine assistant in child care and an office-based telemedicine clinician. Devices used were an all-purpose digital camera (with attachments designed to facilitate capture of ear, nose, throat, skin, and eye images) and an electronic stethoscope. ADI indexed illness that had interrupted care and education for children and burdened both parents and the community with work loss and health care-related costs. Detailed attendance records and staff and parent interviews provided data. The total number of days of attendance expected from all registered children over the course of a week (total child-days) served as the denominator in calculating rates for ADI. The center-week served as the primary unit of analysis. This study is descriptive in character; statistics are not inferential but instead serve to summarize observations.
Results. For the 400 weeks of valid observations contributed by the 5 centers, the mean ADI was 6.41 absences per 100 child-days per week. In bivariate analysis, predictors of ADI were children's mean age, child care center, proportion of children covered by Medicaid, season of the year, and availability of telemedicine. ADI during weeks with telemedicine (4.07 absences per 100 child-days) was less than half that during weeks without telemedicine (8.78 absences per 100 child-days). After adjusting for potentially confounding variables using the generalized estimating equations method, telemedicine remained the strongest predictor of ADI. A 63% reduction in ADI was attributable to telemedicine, an effect similar to the 59% variation in ADI with season of the year. During the 201 total weeks that telemedicine services were available, 940 telemedicine encounters occurred. Telemedicine clinicians for these 940 encounters recommended exclusion from child care for 7.0% and in-person visits for 2.8% of the children. In surveys, parents indicated that 91.2% of telemedicine contacts allowed them to stay at work and that 93.8% of problems managed by telemedicine would otherwise have led to an office or emergency department visit.
Conclusions. Telemedicine holds substantial potential to reduce the impact of illness on health and education of children, on time lost from work in parents, and on absenteeism in the economy.
Key Words: access to health care child care health services research illness telemedicine
Abbreviations: ADI, absence from child care due to illness df, degrees of freedom
Care outside the home has become the norm for preschool children in the United States. In 1995, 60% of children from birth to 5 years of age participated in a nonparental child care or early education program such as Head Start.1 With continuation of the trend for young mothers to join the workforce and the advent of welfare-to-work programs throughout the United States, this proportion is undoubtedly larger today.
Acute illness is a very common and difficult problem for everyone involved in preschool-aged child care. Higher incidence and greater severity of illness among children in child care than among children in home care is well documented,2–5 although evidence suggests that exposure to child care centers is associated with lower rates of subsequent illness.6,7 Substantial economic and social costs arise from health care for affected children, parents' time lost from work, school, and family responsibilities, and children's time lost from educational programs.8–11
Lost parent time impacts industry through diminished worker productivity and may threaten employment,12 especially for entry-level workers. One report found that a child's illness accounted for 40% of missed work for parents using child care.9 Another study based on a nationally representative sample of working women found that only 39% had someone they could call on to help with child care the next time their child is sick.13 Most women reported that they either would need to miss work (49%) or wouldn't know what to do (7%) when this occurs. Work absence to care for a sick child means loss of pay for most women of lower socioeconomic status.13
Spanning distances too great and transcending times too short, telemedicine holds the potential to surmount major access barriers in cities as well as rural areas. Telemedicine has been used to overcome constraints of space and time that impede access to children's health care in both rural and urban schools.14–17 Important applications to subspecialty care for chronic problems in rural areas have been reported.18,19 Based on evidence supporting reliability, efficacy, and feasibility of a telemedicine model designed for diagnosis and treatment of common, acute, childhood illness, the Health-e-Access demonstration project in Rochester, New York, began a telemedicine service involving inner-city child care centers starting in May 2001. Because of the importance of child care to a large proportion of families worldwide and because of the health, educational, and economic implications of absence from child care due to illness (ADI), the objective of the present analysis was to assess the impact of this telemedicine model on ADI in the first 5 child care centers enrolled in our program.
METHODS
Design and Setting
A before-and-after comparison with historical and concurrent controls was used to study the impact of the Health-e-Access telemedicine program in 5 child care centers in Rochester, New York. Figure 1 displays study design and execution. Observations occurred between January 1, 2001, and June 30, 2003. Participating centers were required to collect baseline attendance data for a minimum of 18 weeks before introduction of telemedicine. Telemedicine had not yet begun in the fifth center at the time that the observation period ended. Consequently, this center contributed only "before" data.
Child care centers were located in impoverished inner-city areas. The striking contrast in health indicators for children dwelling in these areas to those for children dwelling in Rochester's relatively affluent suburbs have been reported previously.20–22 Centers were chosen for participation based on recommendations of local child advocacy organizations, interest of center leadership, and center resources to comply with program requirements. The first 3 centers to participate were required to contribute the time of the telehealth assistants and the time required for gathering attendance information. Head Start programs operated in all centers except center 2. Center 4 operated exclusively as a Head Start program. Medicaid covered 66% of children enrolled, and the county Department of Social Services subsidized child care for >95%. Average total enrollment of all 5 centers together was 693 children, representing 6% of all children receiving subsidized child care in Monroe County, New York, in 2003.23
Parents signed for consent to allow reporting of attendance, illness, and health services utilization for study purposes. They subsequently signed a second consent for participation in telemedicine service. Before any telemedicine contact, parents were informed about health concerns and asked whether they wished that their child be evaluated through telemedicine. Centers tracked attendance and ADI for all children registered, and they provided anonymous records of this information for children who had not yet signed the first consent.
Telemedicine Model
The telemedicine model was designed to enable diagnosis and treatment decisions for acute problems that commonly arise in child care settings. In addition to an ill child, participants in telemedicine encounters at the child care site included a telemedicine assistant and sometimes a parent. Parent involvement usually occurred when the parent identified a problem at drop-off time or when the parent was employed at the child care center. These individuals interacted with a telemedicine clinician located at the Golisano Children's Hospital of the University of Rochester Medical Center.
The service used commercially available computer equipment and telemedicine peripheral devices and broadband connections. Technical specifications are available from us on request. Sufficient bandwidth was available to enable clear, synchronous visual and audio communication in real time. The child care site included a simple teleconferencing camera, an all-purpose digital camera with attachments designed to facilitate capture of ear, nose, throat, skin, and eye images, and an electronic stethoscope.
Child care center staff, trained to function as telemedicine assistants, included individuals who had no prior health care training or experience as well as individuals who were certified nurse assistants and licensed practical nurses. Training focused on symptom-driven protocols and electronic forms that guided collection of present and past medical history and use of telemedicine peripheral devices. Telemedicine assistants learned to obtain diagnostic-quality fixed images and video clips (eg, tympanic membrane, skin, throat) and electronic stethoscope audio files. Training involved 40 hours of didactic teaching by the Health-e-Access program coordinator and its nurse manager. Obtaining vital signs, removing cerumen, rapid streptococcal antigen testing, and administration of nebulized medication comprised additional skills addressed. Guidance by the telemedicine clinicians during telemedicine visits served to enhance and solidify skills. Additional details on training are available from us on request.
Guidance to telemedicine assistants, child care staff, and parents was that this service could be considered for any acute health problem that they viewed as important, especially those that might require exclusion from child care until medical evaluation was obtained. We emphasized that there were limitations in the scope of problems that could be addressed appropriately and that telemedicine clinicians would refer children to a higher level of care, after contacting their primary care provider, if they were not confident about the appropriateness of diagnosis or treatment.
Telemedicine participants connected for real-time interaction at an appointed time, usually within 30 minutes after a consultation request to the clinician by the telehealth assistant. Evaluations generally began with review of the forwarded information (eg, images, audio clips, text) by the clinician. Sometimes, no additional information was required to complete the consultation, ie, to make diagnostic and treatment decisions with confidence. Real-time and store-and-forward information exchange between child care and clinician sites was incorporated in the model because of the belief that visual assessment of the child's overall appearance, respiratory pattern, and behavior was essential to accurate assessment of many acute problems in preschool children. Real-time interaction also enabled clinicians and telehealth assistants to work efficiently together as needed to amplify medical history and improve the diagnostic quality of fixed images, video files, or audio files. The clinician also guided the telehealth assistant in real time to elicit information that required tactile sensation, such as palpable attributes of skin lesions and tenderness of lymph nodes. Finally, expected and apparent benefits of the real-time interface included enhanced communication and increased acceptability to clinicians, parents, and child care staff.
Telehealth clinicians included 2 experienced general pediatricians (K.M.M. and N.E.H.) as well as a pediatric nurse practitioner. The nurse practitioner had no prior experience with telemedicine. Preparation for telemedicine by the pediatricians included a total of 90 telemedicine encounters performed as part of a study of the reliability and efficacy of telemedicine for acute pediatric problems. The nurse practitioner conducted telemedicine visits, consulting one of the pediatricians as needed. Clinicians offered service between 7:30 AM and 4:30 PM, Monday through Friday. When indicated, prescriptions were called into a local pharmacy. Arrangements were made with 1 pharmacy to deliver medications to child care centers, an initiative that further promoted early intervention and enhanced convenience for families. After visits, information on diagnosis and treatment recommendations for parents was faxed to child care centers. Telehealth assistants discussed this information, supplemented by standard handouts on common acute illness, with parents. During clinical situations in which diagnosis or management was not straightforward, discussion with the telehealth clinician via phone or telemedicine link also occurred. A summary of the telemedicine visit was also faxed to the child's regular primary care physician. For evolving and potentially serious problems as well as problems beyond the clinical scope of this model, the telehealth clinician called the primary care physician's office to facilitate appropriate subsequent care.
Measurement
The organizational level at which the intervention was implemented was the individual child care center. Although the number of children registered at each center was relatively stable, the children enrolled in a center changed frequently because of change in parent employment status and work schedule. Enrollment periods generally began and ended with the beginning and ending of a week. For these reasons, the natural unit of analysis for ADI was the center-week, defined as the aggregate of all observations for a center for the week. To standardize observations across centers, rates for ADI were calculated as days absent per 100 child-days of expected attendance per week. A child who was enrolled to attend 5 days per week contributed 5 child-days to this measure. Total center enrollment, regardless of telemedicine program participation, was used in calculating rates of ADI.
ADI was based on attendance data gathered every day that child care centers operated. Attendance logs were completed daily for each child care class and faxed or e-mailed to the study data center. The staff classified the reason for absence as illness-related or not. For ADI, health care utilization was recorded also. Research staff reviewed attendance information daily to ensure completeness. If a child was sent home from child care as a result of illness, this event was counted as a half day of absence. Time missed from child care for health maintenance care or routine office visits for chronic problems was not considered illness-related absence.
Interview data about the importance of illness and telemedicine service was obtained for each family. After the first telemedicine visit for a child, a parent was contacted by phone and queried about who requested the visit (parent or child care staff), impact on health care use and time lost from work or school, importance of telemedicine in choosing among child care alternatives, and follow-up health services use for this illness. Multiple attempts were made to contact families at various times of day and on various days of the week. Attempts to contact families after their first use of telemedicine were continued until 100 families were interviewed from each child care center.
Several variables with known or potential relationship to illness rates and health care use for illness among children in a center were also collected because of their potential to confound a relationship between telemedicine and ADI. Potentially confounding variables encompassed time of year, center, average socioeconomic status, health insurance, and age of each center's children. Time-of-year variables were 4-week blocks and illness seasons. Thirteen 4-week blocks were used to provide periods of equal duration. These 13 periods were aggregated into 2 periods representing seasons with low and high rates of illness. The proportion of children covered by Medicaid was used as both an indicator of socioeconomic status of a center's population and type of health insurance. Whereas Medicaid covered the majority of children, almost none lacked health insurance. The proportion of children in each center covered by Medicaid was based on all children registered in the center for each week of observation. Mean age of registered children was used to assess the effect of age on ADI.
Analysis
Because ADI rate was a continuous variable, bivariate relationships between potentially confounding variables and ADI were first assessed with Student's t tests or 1-way analysis of variance for categorical variables and with correlation analysis for other continuous variables.
Centers varied substantially in size, in availability of telehealth assistants, and in events occurring over the course of the study period, such as equipment failures. Thus, weekly ADI data were likely correlated within a center. We therefore used methods for analyzing longitudinal data. The weekly ADI rate was modeled by using Poisson regression. The number of ADIs in a given week was assumed to follow a Poisson distribution, with the number of person-days at risk each week serving as the denominator. The natural log of the true ADI rate was then assumed to be linear in the covariates. In determining which covariates to include, we considered potential confounding factors. Each center had weeks without telemedicine, and all but 1 had weeks with telemedicine. In addition, centers entered the study at different dates. In the regression model, we therefore attempted to isolate the treatment effect from seasonal effects and from the effect of new centers entering the study, as well as from differing characteristics of the centers. An indicator for telemedicine was included in the model and was the covariate of primary interest. The average age of registered children and the proportion of children covered by Medicaid were center-level time-varying covariates that were included in the model as linear predictors. To account for yearly and seasonal differences in ADI rate, we included indicators for year (2002 and 2003, with 2001 as the referent category) and season (low-rate season as referent category) in the model, as well as interactions between year and season. Carefully accounting for time trends was crucial. Otherwise, if the introduction of telemedicine coincided with a seasonal change in ADI rate, the estimated telemedicine effect would capture both seasonal and treatment differences. Finally, parameters were estimated by using the generalized estimating equations method24 and implemented by using SAS PROC GENMOD (SAS Institute, Cary, NC). This method adjusts standard errors to account for within-center correlation.
RESULTS
Enrollment and Observations
The 5 centers contributed a total of 423 weeks of observations during the 2.5 years studied. The first 3 centers contributed most weeks. Distribution of observations among the centers and the status of telemedicine service over this time are detailed in Fig 1. Centers are identified by numbers indicating the order in which telemedicine was initiated. A minimum of 18 weeks of baseline attendance data were required before initiation of telemedicine. Eight of these weeks were required to fall during the months of November through March, because these months have the highest incidence of acute illness. Factors beyond the control of investigators also influenced the pattern of observations substantially. Observations were begun in 3 centers with pilot funding. Additional funding allowed observations in additional centers beginning in November 2002. Observations for 3 weeks encompassing the end and beginning of each calendar year were excluded from final analysis. Days of operation during these weeks varied substantially among centers and years, and attendance varied highly on days that child care was available. Excluding the 23 weeks of holiday observations from final analysis left 400 weeks for analysis. Telemedicine was offered for 201 of these 400 weeks. Among the 199 weeks with no telemedicine service, 177 weeks preceded telemedicine start-up. Lack of service was a consequence of equipment failure for 20 weeks and absence of a center's telemedicine assistant for 2 weeks.
Telemedicine was initiated in a staggered fashion, because that pattern allowed concurrent comparisons with centers lacking telemedicine and it allowed training personnel and technical support in 1 center at a time. Administrative obstacles (eg, with purchasing procedures) sometimes delayed projected start dates. Center 3 chose to be the last to initiate telemedicine among the 3 centers covered by initial funding because of temporary space limitation caused by renovations. Technical problems or lack of a telehealth assistant forced suspension of telemedicine during several periods for the first 3 centers, most notably for 21 center-weeks during the second winter of operation. This interruption of service was caused by modification of the medical center's data firewall, which inadvertently precluded connectivity. Center-weeks with suspended telemedicine service were grouped in analysis with other weeks (baseline) without telemedicine service. Adequate attendance data were not available for center 1 in July and August 2002 because of a staffing problem. Although still voicing enthusiasm about the value of telemedicine, center 2 terminated telemedicine service and attendance data collection after the first week in May 2003. Center 2 administrators made this decision because they had difficulty allocating adequate staff time to telemedicine responsibilities and had fulfilled their commitment of staff time as "in-kind" matching to grant funds. This relatively small center (average registration of 68 children), which maintained a smaller staff and thus had less flexibility in staffing, made only limited use of telemedicine over the 4 months preceding termination (8 telemedicine encounters versus 19 encounters for the same 4-month period in the previous year). Center 1 had fulfilled its commitment of "in-kind" contribution of staff time and ceased collection of attendance data at the end of May 2003, although telemedicine service continued despite lack of grant funds to support the telemedicine assistant. Valid data on the number of registered children was not available for center 4 at the end of June 2003 because of the transition in Head Start programs at the end of the school year.
Total enrollment averaged 138.6 children per center-week and ranged between 68.2 and 335.8 children for the 5 participating centers. Telemedicine participation increased substantially over time after the service became available. Rates of telemedicine participation (Table 1), based on the proportion of enrolled children with consent for telemedicine service, was substantially greater in later weeks (75.7%) than it was for centers in the first 16 weeks (57.1%) of telemedicine availability. The primary factor limiting participation was the logistic challenge of obtaining consent, particularly for children attending child care for only for a short period of time and for children participating in programs in which parents were not responsible for drop-off and pick-up of children. Reluctance to use telemedicine and, especially, to participate in any study were commonly cited by child care staff as barriers to participation, but these considerations seemed unimportant to parents as soon as they were confronted with a specific illness experience in which telemedicine might be used.
Absence From Child Care
The programwide mean ADI over all weeks observed was 6.31 absences per week per 100 children enrolled (Table 1). On average, 6.28 children accounted for these 6.31 absences. The rate for overall absences was substantially greater: 91.0 absences per 100 children enrolled. In addition to ADI, reasons for absences overall included a broad range such as variation in a parent's work schedule, loss of employment, parent illness, child missed the Head Start bus, and other transportation problems.
Predicting ADI
As indicated in Table 1, ADI was much less (t = 7.61; degrees of freedom [df] = 322; P < .0001) during weeks with telemedicine service (4.06 per 100 children) than during weeks without it (8.78 per 100 children). In addition, ADI varied substantially among centers (Table 1) and with time of year (Table 2). Figure 2 displays ADI over time by center. Reduction in ADI after the initiation of telemedicine, most apparent in Fig 2 for centers 1 and 3, seems less dramatic for periods 2 through 7 in 2003, which corresponds to periods in which telemedicine clinicians were unable to respond as quickly to requests from child care centers for visits because of illness in the telemedicine nurse practitioner.
The 52 weeks of the year were aggregated to form a dichotomous variable reflecting 2 levels for rates of illness: high and low. Definition of illness seasons (illness, low illness) was based on several decades of pediatric infectious disease surveillance in this community,25–27 observations on variation in acute illness hospitalization rates,28 and study observations during weeks without telemedicine (Table 2). The illness season included 4-week periods 1 to 3 and 10 to 13 (Fig 1). As shown in Table 2, this dichotomous illness-season variable was a strong predictor of ADI overall (4.81 vs 7.37; t = 4.23; df = 396; P < .0001) and for the 199 center-months without telemedicine service (6.08 vs 10.25; t = 4.39; df = 191; P < .0001). For the 201 months with telemedicine service, however, there was no longer a difference in ADI attributable to season (3.71 vs 4.30; t = 0.96; df = 199; P = .36).
Mean age (4.0 years programwide) and proportion of children covered by Medicaid (66% programwide) both varied substantially by center (see Table 1; P < .001). These variables also held statistically significant relationships with ADI, although size of the correlations was modest. As expected, weeks in which the average age of the children at a center is older were less likely to have an ADI (r = –0.14; P = .006). Center-weeks in which the proportion of children covered by Medicaid was greater had less ADI (r = –0.11; P = .034).
Multivariate Analysis
Results from fitting the Poisson regression model using the generalized estimating equations method are reported in Table 3. All the parameter estimates were statistically significant except for the effect of insurance status. The parameters can be interpreted as log relative risks.29 Thus, after adjusting for age, insurance status, yearly trends, and seasonal trends, the risk of ADI with telemedicine relative to the risk of ADI without telemedicine is calculated as the exponent of the estimate (–0.99) for this parameter, or 0.37. In other words, assuming all confounders have been accounted for in the model, the introduction of telemedicine was associated with a reduction in the risk of ADI to a level that is 0.37 times the risk that prevailed before telemedicine, which represents a 63% reduction in the risk of ADI. As expected, centers with older children on average had a lower risk of ADI (relative risk: 0.66).
Yearly and seasonal trends can be summarized as follows. The ADI rate went up each of the 3 years. In each year except for 2003, the rates were higher in the season that is expected to have a high rate of illness. In 2003, the rates in the low- and high-rate illness seasons were nearly identical. The relative risk associated with a shift from the high- to low-illness seasons, 0.41 (the inverse of the relative risk for season in Table 3 [2.41]), represents a 59% reduction in ADI.
We recognized that a potential limitation with the analysis is that seasonal differences in ADI rate may not be captured by only allowing 2 different rates (ie, illness seasons) per year. To investigate this potential limitation, we also fitted a model that allowed monthly trends by including indicators for month, year, and interactions between month and year in the model. The estimated effect of telemedicine generated from this model was almost the same as that from the simpler model.
All relationships found in multivariate analysis were replicated when analyses were repeated with the data set including holiday weeks.
Telemedicine Visits and Parent Satisfaction
During the 201 weeks that telemedicine service was offered, 940 telemedicine contacts occurred. These 940 contacts involved 362 children and 301 families. Mean age of the children when seen was 2.9 years (SD: 2.0). Traditional Medicaid or Medicaid managed care covered 69.3% of these 362 children, and 1.6% of them were uninsured. Recommendations generated from these 940 evaluations included exclusion from child care for 7.0% and an in-person visit for 2.8%.
Parent-survey responses were obtained from 229 parents, representing 76.1% of the 301 families who had a child evaluated through telemedicine. Among these respondents, 91.2% indicated that the telemedicine contact allowed them to stay at work, with the amount of time saved estimated on average at 4.5 hours (SD: 2.2) per telemedicine visit. In addition, 93.8% of the 227 parents indicated that the problem managed by telemedicine would otherwise have led to an office or emergency department visit, 55.1% indicated that they had brought the problem to the attention of child care staff and requested a telemedicine encounter, and 93.8% indicated that they would choose a child care center with telemedicine over one without this service.
DISCUSSION
Limitations
For both methodologic and substantive reasons, the estimate generated from this study for impact of telemedicine on ADI probably is conservative. We used total enrollment, not just families who had signed consent to participate, as the denominator in calculating rates of ADI. This provided a conservative measure for impact of telemedicine, because telemedicine was available on average to only the 57.7% of families who consented to the use of telemedicine. Also, center 2 made only limited use of telemedicine over the 4 months preceding the end of their participation. The weeks of observation contributed by this center were analyzed as if telemedicine service was operating fully, further contributing to the conservative bias in analysis. This conservative approach was justified in part because of the possibility that change in a center's approach to management of illness resulting from telemedicine might influence ADI beyond the children served by telemedicine.
Reported observations on absence do not indicate how many individual children account for absences. Five absences per 100 children in a typical work-week might be accounted for by as few as 1 and as many as 5 different children. To study the occurrence of illness episodes, rate of health care or prescription use, or incidence of specific diagnoses, the use of center-week as the unit of analysis would be inappropriate. Conclusions about effects on the individual level are central for those studies, and it is not possible to draw direct conclusions about individual-level effects from the present center-based design. Although such studies are relevant to the use of telemedicine in child care, the focus of the present analysis was on the potential for community impact. From the community's perspective, parental work absence is important regardless of the ratio of parents to days of work absence. Analysis indicated that the ratio of children to ADI in this study was almost 1 to 1. On average, 6.28 different children accounted for 6.33 ADI.
The observation of a 63% reduction in ADI may underestimate the potential for impact, because complex innovations do not mature in 2.5 years. Acceptance of innovations takes time. Acceptance of telemedicine by parents, child care staff, and nurses from the community nursing agency that oversees health-related issues in child care for this community was understandably gradual. Grant funding did not provide resources to disseminate information about Health-e-Access. "Marketing" this innovation relied heavily on word of mouth. It also takes time to optimize efficiency. Telemedicine assistants and clinicians were new participants in a new system. They mostly learned together.
Human factors are critical to the effective use of telemedicine, as with use of any tool. Achievements of this program reflect the good will, collaborative spirit, creativity, and flexibility of many individuals within a broad range of organizations. In addition to the potential of telemedicine, achievements reflect inputs of child care leadership, child care staff, parents, community physicians on the Health-e-Access Professional Advisory Board, the community nursing agency, the collaborating local broadband telecommunications vendor, information services units within our medical center, telemedicine clinicians, and telehealth assistants. Not every community will develop a system that develops as rapidly or as well as this one. In part because of grant funding, Health-e-Access was able to provide most telemedicine visits within 30 minutes of the time the visit was requested. Whether the ability to respond this rapidly is critical to success is unknown. Other systems may work more effectively. Our experience suggests that telemedicine is a tool that enables collaboration among disparate groups with incentives aligned in the care of children.
Speculation on Mechanism
Although by design this study made no attempt to identify mechanisms for reducing ADI, we believe that several distinct but interrelated attributes of the Health-e-Access telemedicine model were important to achieving this effect. Key attributes included successful completion of most visits, early diagnosis and treatment, on-site health services, safety certification, and developing a culture of collaboration and trust relating to illness. The term "visit completion" was used for the ability of the clinician to evaluate the child and feel confidant that the information obtained was adequate to make accurate diagnostic decisions and appropriate treatment plans. A telemedicine service without this capacity would be nothing more than an expensive telephone-triage system.
On-site health service enabled prompt diagnosis and treatment. Prevailing child care program policies require medical evaluation before the return of children excluded from care because of illness. Thus, parents usually bring excluded children to medical attention relatively soon. However, parents cannot accomplish this as rapidly as medical evaluation was achieved through Health-e-Access. In this program, evaluation almost always was completed within 1 hour of the time that the problem was identified. In the typical child care setting, a child awaking febrile from a midday nap prompts an immediate call for the parent to remove their child and then, for families who have a primary care physician, a call to that physician's office.
Based on telephone assessment, the primary care physician's office often will recommend a next-day visit. Although usually appropriate from a medical perspective, a next-day visit may not be desirable from the parent's perspective. Next-day visits often require loss of parent time from work. Waiting until the next day may prolong parental anxiety and children's discomfort. Yet, same-day alternatives to the office visit (an after-hours clinic or emergency department visit) are more costly, time consuming, inconvenient, and often demeaning.
No usual alternative readily resolves problems posed to parents by the dreaded call from child care, but on-site telemedicine usually does. Even when the telemedicine visit leads to exclusion from child care, use of usual health services is generally avoided. That by itself may reduce some ADI and some work loss. In our experience, however, most often the telemedicine clinician both completed the visit and provided reassurance that exclusion from child care was not required.
Safety certification played a substantial role in reducing ADI. We defined safety certification as the clinician's judgment that exclusion from child care offered no substantial benefit to the child with a health problem, child care staff, or other children. We believe that the telemedicine link was necessary, although not sufficient, for safety certification. The necessary contribution of telemedicine was to bring clinical acumen on a case-by-case basis to decisions about exclusion. However, safety certification also rested on judgments about guidelines regarding exclusion from child care because of illness. It is appropriate that many guidelines promulgated by the American Academy of Pediatrics (such as Red Book30) and the American Public Health Association31 call for exclusion from day care until the child is evaluated by a clinician. In our view, justification for Red Book exclusion criteria often rests on the value of clinician evaluation in excluding more serious conditions; exclusion forces parents to have their child evaluated by a clinician. Telemedicine ensures that this evaluation occurs on site without requiring exclusion (guidelines for exclusion used in the Health-e-Access program, which reflect that capacity of the telemedicine clinician to provide safety certification, are available from us on request).
Child care staff and community nurses accepted exclusion guidelines (as modified by Health-e-Access leadership) hesitantly and after several discussions. Parents and community-based physicians accepted modified exclusion guidelines more readily, as expected based on prior research32,33 and on focus groups we conducted in preparation for this program. In the usual child care setting, exclusion because of illness remains a contentious issue. Parents often feel that child care staff are unreasonable, inconsistent, and guilty of favoritism in their application of guidelines. The staff often feels that parents are deceitful and irresponsible in their attempts to subvert guidelines.
Safety certification played a central role in replacing contention with collaboration and trust in addressing illness. Telemedicine usually enabled child care to become the source of the solution rather than the cause of the quandary. If parents conveyed health concerns to staff and facilitated phone access while at work, Health-e-Access usually allowed telemedicine assistants to solve health problems whenever they arose. Additionally, telemedicine assistants ensured parents' understanding of diagnosis and treatment recommendations. Even when exclusion was an outcome, Health-e-Access dissipated distrust by creating the opportunity for parents and staff to defer to the judgment of a medical authority.
Rationale for Reimbursement
Although Medicaid provides reimbursement for some form of telemedicine in at least 18 states,34 in most states Medicaid does not reimburse for telemedicine service. The value of safety certification in child care should be considered in cost-effectiveness analysis of Medicaid reimbursement for telemedicine. The conundrum of contention currently surrounding illness in child care is the consequence of a public-sector catch-22 that entraps child care staff and parents alike. Both are caught in an unintended conflict between public health and social service missions of county government. Welfare-to-work programs administered by social service agencies seek to reduce dependency and enhance economic well-being. Inner-city parents need reliable child care in responding to these initiatives. In guarding the public's health, health departments promote guidelines for exclusion that cast a broad net, ensuring that all serious problems are detected. It seems, unfortunately, that guidelines for use by nonprofessionals also exclude a large number of mildly ill children who need not be, which reduces the effectiveness of welfare-to-work programs. Results of this study suggest that yields from government investment in welfare-to-work programs (in subsidizing child care) and community health might be substantially greater given modest additional investment in supporting telemedicine service that emulates the model studied.
CONCLUSIONS
Telemedicine holds substantial potential to reduce the impact of illness on children and families using child care. Availability of Health-e-Access was associated with a marked reduction of ADI. Direct measurement of the effects of reduced ADI on child health and education, family functioning, health care use, and worker productivity was beyond the scope of available resources. Substantial impact of the Health-e-Access model in all these areas is plausible, was supported by parent-survey results, and seems likely to extend across the socioeconomic spectrum.
ACKNOWLEDGMENTS
We are grateful for funding and in-kind contributions provided by the US Department of Commerce Technology Opportunities Program, Robert Wood Johnson Foundation Local Initiative Funding Partners Program, 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, Frontier Telecommunications Corporation, anonymous donors, Volunteers of America Children's Center, Carlson Metro YMCA Children's Center, Generations Child Care, ABC Austin Street Head Start Program, and Lewis Street YMCA Child Care.
FOOTNOTES
Accepted Sep 14, 2004.
Conflict of interest: Dr McConnochie, Mrs Wood, and Dr Herendeen hold equity positions in Tel-e-Atrics.
REFERENCES
West J, Wright D, Hausken EG. Child Care and Early Education Program Participation of Infants, Toddlers, and Preschoolers. Washington, DC: US Department of Education; 1995
Wald ER, Dashefsy B. Frequency and severity of infections in day care. J Pediatr. 1988;112 :540 –546
Bartlett AV, Moore M, Gary GW, Starko KM, Erben JJ, Meredith BA. Diarrheal illness among infants and toddlers in day care centers. II. Comparison with day care homes and households. J Pediatr. 1985;107 :503 –509
Fleming DW, Cochi SL, Hightower AW, Broome CV. Childhood upper respiratory tract infections: to what degree is incidence affected by day-care attendance. Pediatrics. 1987;79 :55 –60
Hurwitz ES, Gunn WJ, Pinsky PF, Schonberger LB. Risk of respiratory illness associated with day-care attendance: a nationwide study. Pediatrics. 1991;87 :62 –69
Ball TM, Holberg CJ, Aldous MB, Martinez FD, Wright AL. Influence of attendance at day care on the common cold from birth through 13 years of age. Arch Pediatr Adolesc Med. 2002;156 :121 –126
Kramer U, Heinrich J, Wjst M, Wichmann HE. Age of entry to day nursery and allergy in later childhood. Lancet. 1999;353 :450 –454
Carabin H, Gyorkos TW, Soto JC, Penrod J, Joseph L, Collet JP. Estimation of direct and indirect costs because of common infections in toddlers attending day care centers. Pediatrics. 1999;103 :556 –564
Bell DM, Gleiber DW, Mercer AA, et al. Illness associated with child day care: a study of incidence and cost Comparison with day care homes and households. Am J Public Health. 1989;79 :479 –484
Nurmi T, Salminen E, Ponka A. Infections and other illnesses of children in day-care centers in Helsinki. II: The economic losses. Infection. 1991;19 :331 –335
Jordan AE. The unresolved child care dilemma: care for the acutely ill child. Rev Infect Dis. 1986;8 :626 –630
Doolittle GC, Williams AR, Cook DJ. An estimation of costs of a pediatric telemedicine practice in public schools. Med Care. 2003;41 :100 –109
Whitten P, Kingsley C, Cook DJ, Swirczynski D, Doolittle GC. School-based telehealth: an empirical analysis of teacher, nurse, and administrator perceptions. J Sch Health. 2001;71 :173 –199
Young TL, Ireson C. Effectiveness of school-based telehealth care in urban and rural elementary schools. Pediatrics. 2003;112 :1088 –1094
Whitten P, Cook DJ, Shaw P, Ermer D, Goodwin J. Telekid Care: bringing health care into schools. Telemed J. 1998;4 :335 –343
Marcin JP, Ellis J, Mawis R, Nagrampa E, Nesbitt TS, Dimand RJ. Using telemedicine to provide pediatric subspecialty care to children with special health care needs in an underserved rural community. Pediatrics. 2004;113 :1 –6
Lobe TE. Telemedicine and the future of healthcare for our children [commentary]. Pediatrics. 2004;113 :130
McConnochie KM, Roghmann KJ, Liptak GS. Socioeconomic variation in discretionary and mandatory hospitalization of infants: an ecologic analysis. Pediatrics. 1997;99 :774 –784
McConnochie KM, Roghmann KJ, Liptak GS. Variation in hospitalization rates for infants within a community: diagnostic clusters as indexes of community health. Eval Health Prof. 1998;21 :332 –361
Goodman J. Child care squeeze: low-income parents worry and scramble to cope with county's shrinking safety net of subsidies. Democrat and Chronicle. February 7, 2004:1
Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73 :13 –22
Hall CB, Douglas RG Jr. Respiratory syncytial virus and influenza. Practical community surveillance. Am J Dis Child. 1976;130 :615 –620
Hall CB, Walsh EE, Schnabel KC, et al. Occurrence of groups A and B of respiratory syncytial virus over 15 years: associated epidemiologic and clinical characteristics in hospitalized and ambulatory children. J Infect Dis. 1990;162 :1283 –1290
Long CE, Hall CB, Cunningham CK, et al. Influenza surveillance in community-dwelling elderly compared with children. Arch Fam Med. 1997;6 :459 –465
McConnochie KM, Roghmann KJ, Liptak GS. Hospitalization for lower respiratory tract illness in infants: variation in rates among counties in New York State and areas in Monroe County. J Pediatr. 1995;126 :220 –229
Diggle PJ, Heagerty P, Liang K-Y, Zeger SL. Analysis of Longitudinal Data. 2nd ed. Oxford, United Kingdom: Oxford University Press; 2002
American Academy of Pediatrics. 2003 Red Book: Report of the Committee on Infectious Diseases. 26th ed. Elk Grove Village, IL: American Academy of Pediatrics; 2003
American Public Health Association, American Academy of Pediatrics. Caring for Our Children. National Health and Safety Performance Standards: Guidelines for Out-of-Home Child Care Programs. Washington, DC: American Public Health Association; 2000
Pappas DE, Schwartz RH, Sheridan MJ, Hayden GF. Medical exclusion of sick children from child care centers: a plea for reconciliation. 2000;93 :575 –578
Landis SE, Earp JL, Sharp M. Day-care center exclusion of sick children: comparison of opinions of day-care staff, working mothers, and pediatricians. Pediatrics. 1988;81 :662 –667
Centers for Medicaid and Medicare Services. States where Medicaid reimbursement of services utilizing telemedicine is available. Available at: www.cms.hhs.gov/states/telelist.asp. Accessed March 12, 2005(Kenneth M. McConnochie, M)
Department of Biostatistics and Computational Biology, University of Rochester Medical Center, Rochester, New York
ABSTRACT
Background. Common acute illness challenges everyone involved in child care. Impoverished inner-city families, whose children are most burdened by morbidity and whose reliance on child care is most important, are those least equipped to deal with this challenge.
Objective. To assess the impact of telemedicine on absence from child care due to illness (ADI).
Design/Methods. A before-and-after design with historical and concurrent controls was used to study ADI in 5 inner-city child care centers in Rochester, New York, between January 1, 2001, and June 30, 2003. Enrollment averaged 138 children per center, of whom Medicaid covered 66%. Center 5 provided only concurrent controls. Telemedicine service began in the first 4 centers in a staggered fashion starting in May 2001. Baseline data on ADI before availability of telemedicine were collected in each center for a minimum of 18 weeks. The telemedicine model for diagnosis and treatment of common acute problems involved both real-time and store-and-forward information exchange between a child and telemedicine assistant in child care and an office-based telemedicine clinician. Devices used were an all-purpose digital camera (with attachments designed to facilitate capture of ear, nose, throat, skin, and eye images) and an electronic stethoscope. ADI indexed illness that had interrupted care and education for children and burdened both parents and the community with work loss and health care-related costs. Detailed attendance records and staff and parent interviews provided data. The total number of days of attendance expected from all registered children over the course of a week (total child-days) served as the denominator in calculating rates for ADI. The center-week served as the primary unit of analysis. This study is descriptive in character; statistics are not inferential but instead serve to summarize observations.
Results. For the 400 weeks of valid observations contributed by the 5 centers, the mean ADI was 6.41 absences per 100 child-days per week. In bivariate analysis, predictors of ADI were children's mean age, child care center, proportion of children covered by Medicaid, season of the year, and availability of telemedicine. ADI during weeks with telemedicine (4.07 absences per 100 child-days) was less than half that during weeks without telemedicine (8.78 absences per 100 child-days). After adjusting for potentially confounding variables using the generalized estimating equations method, telemedicine remained the strongest predictor of ADI. A 63% reduction in ADI was attributable to telemedicine, an effect similar to the 59% variation in ADI with season of the year. During the 201 total weeks that telemedicine services were available, 940 telemedicine encounters occurred. Telemedicine clinicians for these 940 encounters recommended exclusion from child care for 7.0% and in-person visits for 2.8% of the children. In surveys, parents indicated that 91.2% of telemedicine contacts allowed them to stay at work and that 93.8% of problems managed by telemedicine would otherwise have led to an office or emergency department visit.
Conclusions. Telemedicine holds substantial potential to reduce the impact of illness on health and education of children, on time lost from work in parents, and on absenteeism in the economy.
Key Words: access to health care child care health services research illness telemedicine
Abbreviations: ADI, absence from child care due to illness df, degrees of freedom
Care outside the home has become the norm for preschool children in the United States. In 1995, 60% of children from birth to 5 years of age participated in a nonparental child care or early education program such as Head Start.1 With continuation of the trend for young mothers to join the workforce and the advent of welfare-to-work programs throughout the United States, this proportion is undoubtedly larger today.
Acute illness is a very common and difficult problem for everyone involved in preschool-aged child care. Higher incidence and greater severity of illness among children in child care than among children in home care is well documented,2–5 although evidence suggests that exposure to child care centers is associated with lower rates of subsequent illness.6,7 Substantial economic and social costs arise from health care for affected children, parents' time lost from work, school, and family responsibilities, and children's time lost from educational programs.8–11
Lost parent time impacts industry through diminished worker productivity and may threaten employment,12 especially for entry-level workers. One report found that a child's illness accounted for 40% of missed work for parents using child care.9 Another study based on a nationally representative sample of working women found that only 39% had someone they could call on to help with child care the next time their child is sick.13 Most women reported that they either would need to miss work (49%) or wouldn't know what to do (7%) when this occurs. Work absence to care for a sick child means loss of pay for most women of lower socioeconomic status.13
Spanning distances too great and transcending times too short, telemedicine holds the potential to surmount major access barriers in cities as well as rural areas. Telemedicine has been used to overcome constraints of space and time that impede access to children's health care in both rural and urban schools.14–17 Important applications to subspecialty care for chronic problems in rural areas have been reported.18,19 Based on evidence supporting reliability, efficacy, and feasibility of a telemedicine model designed for diagnosis and treatment of common, acute, childhood illness, the Health-e-Access demonstration project in Rochester, New York, began a telemedicine service involving inner-city child care centers starting in May 2001. Because of the importance of child care to a large proportion of families worldwide and because of the health, educational, and economic implications of absence from child care due to illness (ADI), the objective of the present analysis was to assess the impact of this telemedicine model on ADI in the first 5 child care centers enrolled in our program.
METHODS
Design and Setting
A before-and-after comparison with historical and concurrent controls was used to study the impact of the Health-e-Access telemedicine program in 5 child care centers in Rochester, New York. Figure 1 displays study design and execution. Observations occurred between January 1, 2001, and June 30, 2003. Participating centers were required to collect baseline attendance data for a minimum of 18 weeks before introduction of telemedicine. Telemedicine had not yet begun in the fifth center at the time that the observation period ended. Consequently, this center contributed only "before" data.
Child care centers were located in impoverished inner-city areas. The striking contrast in health indicators for children dwelling in these areas to those for children dwelling in Rochester's relatively affluent suburbs have been reported previously.20–22 Centers were chosen for participation based on recommendations of local child advocacy organizations, interest of center leadership, and center resources to comply with program requirements. The first 3 centers to participate were required to contribute the time of the telehealth assistants and the time required for gathering attendance information. Head Start programs operated in all centers except center 2. Center 4 operated exclusively as a Head Start program. Medicaid covered 66% of children enrolled, and the county Department of Social Services subsidized child care for >95%. Average total enrollment of all 5 centers together was 693 children, representing 6% of all children receiving subsidized child care in Monroe County, New York, in 2003.23
Parents signed for consent to allow reporting of attendance, illness, and health services utilization for study purposes. They subsequently signed a second consent for participation in telemedicine service. Before any telemedicine contact, parents were informed about health concerns and asked whether they wished that their child be evaluated through telemedicine. Centers tracked attendance and ADI for all children registered, and they provided anonymous records of this information for children who had not yet signed the first consent.
Telemedicine Model
The telemedicine model was designed to enable diagnosis and treatment decisions for acute problems that commonly arise in child care settings. In addition to an ill child, participants in telemedicine encounters at the child care site included a telemedicine assistant and sometimes a parent. Parent involvement usually occurred when the parent identified a problem at drop-off time or when the parent was employed at the child care center. These individuals interacted with a telemedicine clinician located at the Golisano Children's Hospital of the University of Rochester Medical Center.
The service used commercially available computer equipment and telemedicine peripheral devices and broadband connections. Technical specifications are available from us on request. Sufficient bandwidth was available to enable clear, synchronous visual and audio communication in real time. The child care site included a simple teleconferencing camera, an all-purpose digital camera with attachments designed to facilitate capture of ear, nose, throat, skin, and eye images, and an electronic stethoscope.
Child care center staff, trained to function as telemedicine assistants, included individuals who had no prior health care training or experience as well as individuals who were certified nurse assistants and licensed practical nurses. Training focused on symptom-driven protocols and electronic forms that guided collection of present and past medical history and use of telemedicine peripheral devices. Telemedicine assistants learned to obtain diagnostic-quality fixed images and video clips (eg, tympanic membrane, skin, throat) and electronic stethoscope audio files. Training involved 40 hours of didactic teaching by the Health-e-Access program coordinator and its nurse manager. Obtaining vital signs, removing cerumen, rapid streptococcal antigen testing, and administration of nebulized medication comprised additional skills addressed. Guidance by the telemedicine clinicians during telemedicine visits served to enhance and solidify skills. Additional details on training are available from us on request.
Guidance to telemedicine assistants, child care staff, and parents was that this service could be considered for any acute health problem that they viewed as important, especially those that might require exclusion from child care until medical evaluation was obtained. We emphasized that there were limitations in the scope of problems that could be addressed appropriately and that telemedicine clinicians would refer children to a higher level of care, after contacting their primary care provider, if they were not confident about the appropriateness of diagnosis or treatment.
Telemedicine participants connected for real-time interaction at an appointed time, usually within 30 minutes after a consultation request to the clinician by the telehealth assistant. Evaluations generally began with review of the forwarded information (eg, images, audio clips, text) by the clinician. Sometimes, no additional information was required to complete the consultation, ie, to make diagnostic and treatment decisions with confidence. Real-time and store-and-forward information exchange between child care and clinician sites was incorporated in the model because of the belief that visual assessment of the child's overall appearance, respiratory pattern, and behavior was essential to accurate assessment of many acute problems in preschool children. Real-time interaction also enabled clinicians and telehealth assistants to work efficiently together as needed to amplify medical history and improve the diagnostic quality of fixed images, video files, or audio files. The clinician also guided the telehealth assistant in real time to elicit information that required tactile sensation, such as palpable attributes of skin lesions and tenderness of lymph nodes. Finally, expected and apparent benefits of the real-time interface included enhanced communication and increased acceptability to clinicians, parents, and child care staff.
Telehealth clinicians included 2 experienced general pediatricians (K.M.M. and N.E.H.) as well as a pediatric nurse practitioner. The nurse practitioner had no prior experience with telemedicine. Preparation for telemedicine by the pediatricians included a total of 90 telemedicine encounters performed as part of a study of the reliability and efficacy of telemedicine for acute pediatric problems. The nurse practitioner conducted telemedicine visits, consulting one of the pediatricians as needed. Clinicians offered service between 7:30 AM and 4:30 PM, Monday through Friday. When indicated, prescriptions were called into a local pharmacy. Arrangements were made with 1 pharmacy to deliver medications to child care centers, an initiative that further promoted early intervention and enhanced convenience for families. After visits, information on diagnosis and treatment recommendations for parents was faxed to child care centers. Telehealth assistants discussed this information, supplemented by standard handouts on common acute illness, with parents. During clinical situations in which diagnosis or management was not straightforward, discussion with the telehealth clinician via phone or telemedicine link also occurred. A summary of the telemedicine visit was also faxed to the child's regular primary care physician. For evolving and potentially serious problems as well as problems beyond the clinical scope of this model, the telehealth clinician called the primary care physician's office to facilitate appropriate subsequent care.
Measurement
The organizational level at which the intervention was implemented was the individual child care center. Although the number of children registered at each center was relatively stable, the children enrolled in a center changed frequently because of change in parent employment status and work schedule. Enrollment periods generally began and ended with the beginning and ending of a week. For these reasons, the natural unit of analysis for ADI was the center-week, defined as the aggregate of all observations for a center for the week. To standardize observations across centers, rates for ADI were calculated as days absent per 100 child-days of expected attendance per week. A child who was enrolled to attend 5 days per week contributed 5 child-days to this measure. Total center enrollment, regardless of telemedicine program participation, was used in calculating rates of ADI.
ADI was based on attendance data gathered every day that child care centers operated. Attendance logs were completed daily for each child care class and faxed or e-mailed to the study data center. The staff classified the reason for absence as illness-related or not. For ADI, health care utilization was recorded also. Research staff reviewed attendance information daily to ensure completeness. If a child was sent home from child care as a result of illness, this event was counted as a half day of absence. Time missed from child care for health maintenance care or routine office visits for chronic problems was not considered illness-related absence.
Interview data about the importance of illness and telemedicine service was obtained for each family. After the first telemedicine visit for a child, a parent was contacted by phone and queried about who requested the visit (parent or child care staff), impact on health care use and time lost from work or school, importance of telemedicine in choosing among child care alternatives, and follow-up health services use for this illness. Multiple attempts were made to contact families at various times of day and on various days of the week. Attempts to contact families after their first use of telemedicine were continued until 100 families were interviewed from each child care center.
Several variables with known or potential relationship to illness rates and health care use for illness among children in a center were also collected because of their potential to confound a relationship between telemedicine and ADI. Potentially confounding variables encompassed time of year, center, average socioeconomic status, health insurance, and age of each center's children. Time-of-year variables were 4-week blocks and illness seasons. Thirteen 4-week blocks were used to provide periods of equal duration. These 13 periods were aggregated into 2 periods representing seasons with low and high rates of illness. The proportion of children covered by Medicaid was used as both an indicator of socioeconomic status of a center's population and type of health insurance. Whereas Medicaid covered the majority of children, almost none lacked health insurance. The proportion of children in each center covered by Medicaid was based on all children registered in the center for each week of observation. Mean age of registered children was used to assess the effect of age on ADI.
Analysis
Because ADI rate was a continuous variable, bivariate relationships between potentially confounding variables and ADI were first assessed with Student's t tests or 1-way analysis of variance for categorical variables and with correlation analysis for other continuous variables.
Centers varied substantially in size, in availability of telehealth assistants, and in events occurring over the course of the study period, such as equipment failures. Thus, weekly ADI data were likely correlated within a center. We therefore used methods for analyzing longitudinal data. The weekly ADI rate was modeled by using Poisson regression. The number of ADIs in a given week was assumed to follow a Poisson distribution, with the number of person-days at risk each week serving as the denominator. The natural log of the true ADI rate was then assumed to be linear in the covariates. In determining which covariates to include, we considered potential confounding factors. Each center had weeks without telemedicine, and all but 1 had weeks with telemedicine. In addition, centers entered the study at different dates. In the regression model, we therefore attempted to isolate the treatment effect from seasonal effects and from the effect of new centers entering the study, as well as from differing characteristics of the centers. An indicator for telemedicine was included in the model and was the covariate of primary interest. The average age of registered children and the proportion of children covered by Medicaid were center-level time-varying covariates that were included in the model as linear predictors. To account for yearly and seasonal differences in ADI rate, we included indicators for year (2002 and 2003, with 2001 as the referent category) and season (low-rate season as referent category) in the model, as well as interactions between year and season. Carefully accounting for time trends was crucial. Otherwise, if the introduction of telemedicine coincided with a seasonal change in ADI rate, the estimated telemedicine effect would capture both seasonal and treatment differences. Finally, parameters were estimated by using the generalized estimating equations method24 and implemented by using SAS PROC GENMOD (SAS Institute, Cary, NC). This method adjusts standard errors to account for within-center correlation.
RESULTS
Enrollment and Observations
The 5 centers contributed a total of 423 weeks of observations during the 2.5 years studied. The first 3 centers contributed most weeks. Distribution of observations among the centers and the status of telemedicine service over this time are detailed in Fig 1. Centers are identified by numbers indicating the order in which telemedicine was initiated. A minimum of 18 weeks of baseline attendance data were required before initiation of telemedicine. Eight of these weeks were required to fall during the months of November through March, because these months have the highest incidence of acute illness. Factors beyond the control of investigators also influenced the pattern of observations substantially. Observations were begun in 3 centers with pilot funding. Additional funding allowed observations in additional centers beginning in November 2002. Observations for 3 weeks encompassing the end and beginning of each calendar year were excluded from final analysis. Days of operation during these weeks varied substantially among centers and years, and attendance varied highly on days that child care was available. Excluding the 23 weeks of holiday observations from final analysis left 400 weeks for analysis. Telemedicine was offered for 201 of these 400 weeks. Among the 199 weeks with no telemedicine service, 177 weeks preceded telemedicine start-up. Lack of service was a consequence of equipment failure for 20 weeks and absence of a center's telemedicine assistant for 2 weeks.
Telemedicine was initiated in a staggered fashion, because that pattern allowed concurrent comparisons with centers lacking telemedicine and it allowed training personnel and technical support in 1 center at a time. Administrative obstacles (eg, with purchasing procedures) sometimes delayed projected start dates. Center 3 chose to be the last to initiate telemedicine among the 3 centers covered by initial funding because of temporary space limitation caused by renovations. Technical problems or lack of a telehealth assistant forced suspension of telemedicine during several periods for the first 3 centers, most notably for 21 center-weeks during the second winter of operation. This interruption of service was caused by modification of the medical center's data firewall, which inadvertently precluded connectivity. Center-weeks with suspended telemedicine service were grouped in analysis with other weeks (baseline) without telemedicine service. Adequate attendance data were not available for center 1 in July and August 2002 because of a staffing problem. Although still voicing enthusiasm about the value of telemedicine, center 2 terminated telemedicine service and attendance data collection after the first week in May 2003. Center 2 administrators made this decision because they had difficulty allocating adequate staff time to telemedicine responsibilities and had fulfilled their commitment of staff time as "in-kind" matching to grant funds. This relatively small center (average registration of 68 children), which maintained a smaller staff and thus had less flexibility in staffing, made only limited use of telemedicine over the 4 months preceding termination (8 telemedicine encounters versus 19 encounters for the same 4-month period in the previous year). Center 1 had fulfilled its commitment of "in-kind" contribution of staff time and ceased collection of attendance data at the end of May 2003, although telemedicine service continued despite lack of grant funds to support the telemedicine assistant. Valid data on the number of registered children was not available for center 4 at the end of June 2003 because of the transition in Head Start programs at the end of the school year.
Total enrollment averaged 138.6 children per center-week and ranged between 68.2 and 335.8 children for the 5 participating centers. Telemedicine participation increased substantially over time after the service became available. Rates of telemedicine participation (Table 1), based on the proportion of enrolled children with consent for telemedicine service, was substantially greater in later weeks (75.7%) than it was for centers in the first 16 weeks (57.1%) of telemedicine availability. The primary factor limiting participation was the logistic challenge of obtaining consent, particularly for children attending child care for only for a short period of time and for children participating in programs in which parents were not responsible for drop-off and pick-up of children. Reluctance to use telemedicine and, especially, to participate in any study were commonly cited by child care staff as barriers to participation, but these considerations seemed unimportant to parents as soon as they were confronted with a specific illness experience in which telemedicine might be used.
Absence From Child Care
The programwide mean ADI over all weeks observed was 6.31 absences per week per 100 children enrolled (Table 1). On average, 6.28 children accounted for these 6.31 absences. The rate for overall absences was substantially greater: 91.0 absences per 100 children enrolled. In addition to ADI, reasons for absences overall included a broad range such as variation in a parent's work schedule, loss of employment, parent illness, child missed the Head Start bus, and other transportation problems.
Predicting ADI
As indicated in Table 1, ADI was much less (t = 7.61; degrees of freedom [df] = 322; P < .0001) during weeks with telemedicine service (4.06 per 100 children) than during weeks without it (8.78 per 100 children). In addition, ADI varied substantially among centers (Table 1) and with time of year (Table 2). Figure 2 displays ADI over time by center. Reduction in ADI after the initiation of telemedicine, most apparent in Fig 2 for centers 1 and 3, seems less dramatic for periods 2 through 7 in 2003, which corresponds to periods in which telemedicine clinicians were unable to respond as quickly to requests from child care centers for visits because of illness in the telemedicine nurse practitioner.
The 52 weeks of the year were aggregated to form a dichotomous variable reflecting 2 levels for rates of illness: high and low. Definition of illness seasons (illness, low illness) was based on several decades of pediatric infectious disease surveillance in this community,25–27 observations on variation in acute illness hospitalization rates,28 and study observations during weeks without telemedicine (Table 2). The illness season included 4-week periods 1 to 3 and 10 to 13 (Fig 1). As shown in Table 2, this dichotomous illness-season variable was a strong predictor of ADI overall (4.81 vs 7.37; t = 4.23; df = 396; P < .0001) and for the 199 center-months without telemedicine service (6.08 vs 10.25; t = 4.39; df = 191; P < .0001). For the 201 months with telemedicine service, however, there was no longer a difference in ADI attributable to season (3.71 vs 4.30; t = 0.96; df = 199; P = .36).
Mean age (4.0 years programwide) and proportion of children covered by Medicaid (66% programwide) both varied substantially by center (see Table 1; P < .001). These variables also held statistically significant relationships with ADI, although size of the correlations was modest. As expected, weeks in which the average age of the children at a center is older were less likely to have an ADI (r = –0.14; P = .006). Center-weeks in which the proportion of children covered by Medicaid was greater had less ADI (r = –0.11; P = .034).
Multivariate Analysis
Results from fitting the Poisson regression model using the generalized estimating equations method are reported in Table 3. All the parameter estimates were statistically significant except for the effect of insurance status. The parameters can be interpreted as log relative risks.29 Thus, after adjusting for age, insurance status, yearly trends, and seasonal trends, the risk of ADI with telemedicine relative to the risk of ADI without telemedicine is calculated as the exponent of the estimate (–0.99) for this parameter, or 0.37. In other words, assuming all confounders have been accounted for in the model, the introduction of telemedicine was associated with a reduction in the risk of ADI to a level that is 0.37 times the risk that prevailed before telemedicine, which represents a 63% reduction in the risk of ADI. As expected, centers with older children on average had a lower risk of ADI (relative risk: 0.66).
Yearly and seasonal trends can be summarized as follows. The ADI rate went up each of the 3 years. In each year except for 2003, the rates were higher in the season that is expected to have a high rate of illness. In 2003, the rates in the low- and high-rate illness seasons were nearly identical. The relative risk associated with a shift from the high- to low-illness seasons, 0.41 (the inverse of the relative risk for season in Table 3 [2.41]), represents a 59% reduction in ADI.
We recognized that a potential limitation with the analysis is that seasonal differences in ADI rate may not be captured by only allowing 2 different rates (ie, illness seasons) per year. To investigate this potential limitation, we also fitted a model that allowed monthly trends by including indicators for month, year, and interactions between month and year in the model. The estimated effect of telemedicine generated from this model was almost the same as that from the simpler model.
All relationships found in multivariate analysis were replicated when analyses were repeated with the data set including holiday weeks.
Telemedicine Visits and Parent Satisfaction
During the 201 weeks that telemedicine service was offered, 940 telemedicine contacts occurred. These 940 contacts involved 362 children and 301 families. Mean age of the children when seen was 2.9 years (SD: 2.0). Traditional Medicaid or Medicaid managed care covered 69.3% of these 362 children, and 1.6% of them were uninsured. Recommendations generated from these 940 evaluations included exclusion from child care for 7.0% and an in-person visit for 2.8%.
Parent-survey responses were obtained from 229 parents, representing 76.1% of the 301 families who had a child evaluated through telemedicine. Among these respondents, 91.2% indicated that the telemedicine contact allowed them to stay at work, with the amount of time saved estimated on average at 4.5 hours (SD: 2.2) per telemedicine visit. In addition, 93.8% of the 227 parents indicated that the problem managed by telemedicine would otherwise have led to an office or emergency department visit, 55.1% indicated that they had brought the problem to the attention of child care staff and requested a telemedicine encounter, and 93.8% indicated that they would choose a child care center with telemedicine over one without this service.
DISCUSSION
Limitations
For both methodologic and substantive reasons, the estimate generated from this study for impact of telemedicine on ADI probably is conservative. We used total enrollment, not just families who had signed consent to participate, as the denominator in calculating rates of ADI. This provided a conservative measure for impact of telemedicine, because telemedicine was available on average to only the 57.7% of families who consented to the use of telemedicine. Also, center 2 made only limited use of telemedicine over the 4 months preceding the end of their participation. The weeks of observation contributed by this center were analyzed as if telemedicine service was operating fully, further contributing to the conservative bias in analysis. This conservative approach was justified in part because of the possibility that change in a center's approach to management of illness resulting from telemedicine might influence ADI beyond the children served by telemedicine.
Reported observations on absence do not indicate how many individual children account for absences. Five absences per 100 children in a typical work-week might be accounted for by as few as 1 and as many as 5 different children. To study the occurrence of illness episodes, rate of health care or prescription use, or incidence of specific diagnoses, the use of center-week as the unit of analysis would be inappropriate. Conclusions about effects on the individual level are central for those studies, and it is not possible to draw direct conclusions about individual-level effects from the present center-based design. Although such studies are relevant to the use of telemedicine in child care, the focus of the present analysis was on the potential for community impact. From the community's perspective, parental work absence is important regardless of the ratio of parents to days of work absence. Analysis indicated that the ratio of children to ADI in this study was almost 1 to 1. On average, 6.28 different children accounted for 6.33 ADI.
The observation of a 63% reduction in ADI may underestimate the potential for impact, because complex innovations do not mature in 2.5 years. Acceptance of innovations takes time. Acceptance of telemedicine by parents, child care staff, and nurses from the community nursing agency that oversees health-related issues in child care for this community was understandably gradual. Grant funding did not provide resources to disseminate information about Health-e-Access. "Marketing" this innovation relied heavily on word of mouth. It also takes time to optimize efficiency. Telemedicine assistants and clinicians were new participants in a new system. They mostly learned together.
Human factors are critical to the effective use of telemedicine, as with use of any tool. Achievements of this program reflect the good will, collaborative spirit, creativity, and flexibility of many individuals within a broad range of organizations. In addition to the potential of telemedicine, achievements reflect inputs of child care leadership, child care staff, parents, community physicians on the Health-e-Access Professional Advisory Board, the community nursing agency, the collaborating local broadband telecommunications vendor, information services units within our medical center, telemedicine clinicians, and telehealth assistants. Not every community will develop a system that develops as rapidly or as well as this one. In part because of grant funding, Health-e-Access was able to provide most telemedicine visits within 30 minutes of the time the visit was requested. Whether the ability to respond this rapidly is critical to success is unknown. Other systems may work more effectively. Our experience suggests that telemedicine is a tool that enables collaboration among disparate groups with incentives aligned in the care of children.
Speculation on Mechanism
Although by design this study made no attempt to identify mechanisms for reducing ADI, we believe that several distinct but interrelated attributes of the Health-e-Access telemedicine model were important to achieving this effect. Key attributes included successful completion of most visits, early diagnosis and treatment, on-site health services, safety certification, and developing a culture of collaboration and trust relating to illness. The term "visit completion" was used for the ability of the clinician to evaluate the child and feel confidant that the information obtained was adequate to make accurate diagnostic decisions and appropriate treatment plans. A telemedicine service without this capacity would be nothing more than an expensive telephone-triage system.
On-site health service enabled prompt diagnosis and treatment. Prevailing child care program policies require medical evaluation before the return of children excluded from care because of illness. Thus, parents usually bring excluded children to medical attention relatively soon. However, parents cannot accomplish this as rapidly as medical evaluation was achieved through Health-e-Access. In this program, evaluation almost always was completed within 1 hour of the time that the problem was identified. In the typical child care setting, a child awaking febrile from a midday nap prompts an immediate call for the parent to remove their child and then, for families who have a primary care physician, a call to that physician's office.
Based on telephone assessment, the primary care physician's office often will recommend a next-day visit. Although usually appropriate from a medical perspective, a next-day visit may not be desirable from the parent's perspective. Next-day visits often require loss of parent time from work. Waiting until the next day may prolong parental anxiety and children's discomfort. Yet, same-day alternatives to the office visit (an after-hours clinic or emergency department visit) are more costly, time consuming, inconvenient, and often demeaning.
No usual alternative readily resolves problems posed to parents by the dreaded call from child care, but on-site telemedicine usually does. Even when the telemedicine visit leads to exclusion from child care, use of usual health services is generally avoided. That by itself may reduce some ADI and some work loss. In our experience, however, most often the telemedicine clinician both completed the visit and provided reassurance that exclusion from child care was not required.
Safety certification played a substantial role in reducing ADI. We defined safety certification as the clinician's judgment that exclusion from child care offered no substantial benefit to the child with a health problem, child care staff, or other children. We believe that the telemedicine link was necessary, although not sufficient, for safety certification. The necessary contribution of telemedicine was to bring clinical acumen on a case-by-case basis to decisions about exclusion. However, safety certification also rested on judgments about guidelines regarding exclusion from child care because of illness. It is appropriate that many guidelines promulgated by the American Academy of Pediatrics (such as Red Book30) and the American Public Health Association31 call for exclusion from day care until the child is evaluated by a clinician. In our view, justification for Red Book exclusion criteria often rests on the value of clinician evaluation in excluding more serious conditions; exclusion forces parents to have their child evaluated by a clinician. Telemedicine ensures that this evaluation occurs on site without requiring exclusion (guidelines for exclusion used in the Health-e-Access program, which reflect that capacity of the telemedicine clinician to provide safety certification, are available from us on request).
Child care staff and community nurses accepted exclusion guidelines (as modified by Health-e-Access leadership) hesitantly and after several discussions. Parents and community-based physicians accepted modified exclusion guidelines more readily, as expected based on prior research32,33 and on focus groups we conducted in preparation for this program. In the usual child care setting, exclusion because of illness remains a contentious issue. Parents often feel that child care staff are unreasonable, inconsistent, and guilty of favoritism in their application of guidelines. The staff often feels that parents are deceitful and irresponsible in their attempts to subvert guidelines.
Safety certification played a central role in replacing contention with collaboration and trust in addressing illness. Telemedicine usually enabled child care to become the source of the solution rather than the cause of the quandary. If parents conveyed health concerns to staff and facilitated phone access while at work, Health-e-Access usually allowed telemedicine assistants to solve health problems whenever they arose. Additionally, telemedicine assistants ensured parents' understanding of diagnosis and treatment recommendations. Even when exclusion was an outcome, Health-e-Access dissipated distrust by creating the opportunity for parents and staff to defer to the judgment of a medical authority.
Rationale for Reimbursement
Although Medicaid provides reimbursement for some form of telemedicine in at least 18 states,34 in most states Medicaid does not reimburse for telemedicine service. The value of safety certification in child care should be considered in cost-effectiveness analysis of Medicaid reimbursement for telemedicine. The conundrum of contention currently surrounding illness in child care is the consequence of a public-sector catch-22 that entraps child care staff and parents alike. Both are caught in an unintended conflict between public health and social service missions of county government. Welfare-to-work programs administered by social service agencies seek to reduce dependency and enhance economic well-being. Inner-city parents need reliable child care in responding to these initiatives. In guarding the public's health, health departments promote guidelines for exclusion that cast a broad net, ensuring that all serious problems are detected. It seems, unfortunately, that guidelines for use by nonprofessionals also exclude a large number of mildly ill children who need not be, which reduces the effectiveness of welfare-to-work programs. Results of this study suggest that yields from government investment in welfare-to-work programs (in subsidizing child care) and community health might be substantially greater given modest additional investment in supporting telemedicine service that emulates the model studied.
CONCLUSIONS
Telemedicine holds substantial potential to reduce the impact of illness on children and families using child care. Availability of Health-e-Access was associated with a marked reduction of ADI. Direct measurement of the effects of reduced ADI on child health and education, family functioning, health care use, and worker productivity was beyond the scope of available resources. Substantial impact of the Health-e-Access model in all these areas is plausible, was supported by parent-survey results, and seems likely to extend across the socioeconomic spectrum.
ACKNOWLEDGMENTS
We are grateful for funding and in-kind contributions provided by the US Department of Commerce Technology Opportunities Program, Robert Wood Johnson Foundation Local Initiative Funding Partners Program, 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, Frontier Telecommunications Corporation, anonymous donors, Volunteers of America Children's Center, Carlson Metro YMCA Children's Center, Generations Child Care, ABC Austin Street Head Start Program, and Lewis Street YMCA Child Care.
FOOTNOTES
Accepted Sep 14, 2004.
Conflict of interest: Dr McConnochie, Mrs Wood, and Dr Herendeen hold equity positions in Tel-e-Atrics.
REFERENCES
West J, Wright D, Hausken EG. Child Care and Early Education Program Participation of Infants, Toddlers, and Preschoolers. Washington, DC: US Department of Education; 1995
Wald ER, Dashefsy B. Frequency and severity of infections in day care. J Pediatr. 1988;112 :540 –546
Bartlett AV, Moore M, Gary GW, Starko KM, Erben JJ, Meredith BA. Diarrheal illness among infants and toddlers in day care centers. II. Comparison with day care homes and households. J Pediatr. 1985;107 :503 –509
Fleming DW, Cochi SL, Hightower AW, Broome CV. Childhood upper respiratory tract infections: to what degree is incidence affected by day-care attendance. Pediatrics. 1987;79 :55 –60
Hurwitz ES, Gunn WJ, Pinsky PF, Schonberger LB. Risk of respiratory illness associated with day-care attendance: a nationwide study. Pediatrics. 1991;87 :62 –69
Ball TM, Holberg CJ, Aldous MB, Martinez FD, Wright AL. Influence of attendance at day care on the common cold from birth through 13 years of age. Arch Pediatr Adolesc Med. 2002;156 :121 –126
Kramer U, Heinrich J, Wjst M, Wichmann HE. Age of entry to day nursery and allergy in later childhood. Lancet. 1999;353 :450 –454
Carabin H, Gyorkos TW, Soto JC, Penrod J, Joseph L, Collet JP. Estimation of direct and indirect costs because of common infections in toddlers attending day care centers. Pediatrics. 1999;103 :556 –564
Bell DM, Gleiber DW, Mercer AA, et al. Illness associated with child day care: a study of incidence and cost Comparison with day care homes and households. Am J Public Health. 1989;79 :479 –484
Nurmi T, Salminen E, Ponka A. Infections and other illnesses of children in day-care centers in Helsinki. II: The economic losses. Infection. 1991;19 :331 –335
Jordan AE. The unresolved child care dilemma: care for the acutely ill child. Rev Infect Dis. 1986;8 :626 –630
Doolittle GC, Williams AR, Cook DJ. An estimation of costs of a pediatric telemedicine practice in public schools. Med Care. 2003;41 :100 –109
Whitten P, Kingsley C, Cook DJ, Swirczynski D, Doolittle GC. School-based telehealth: an empirical analysis of teacher, nurse, and administrator perceptions. J Sch Health. 2001;71 :173 –199
Young TL, Ireson C. Effectiveness of school-based telehealth care in urban and rural elementary schools. Pediatrics. 2003;112 :1088 –1094
Whitten P, Cook DJ, Shaw P, Ermer D, Goodwin J. Telekid Care: bringing health care into schools. Telemed J. 1998;4 :335 –343
Marcin JP, Ellis J, Mawis R, Nagrampa E, Nesbitt TS, Dimand RJ. Using telemedicine to provide pediatric subspecialty care to children with special health care needs in an underserved rural community. Pediatrics. 2004;113 :1 –6
Lobe TE. Telemedicine and the future of healthcare for our children [commentary]. Pediatrics. 2004;113 :130
McConnochie KM, Roghmann KJ, Liptak GS. Socioeconomic variation in discretionary and mandatory hospitalization of infants: an ecologic analysis. Pediatrics. 1997;99 :774 –784
McConnochie KM, Roghmann KJ, Liptak GS. Variation in hospitalization rates for infants within a community: diagnostic clusters as indexes of community health. Eval Health Prof. 1998;21 :332 –361
Goodman J. Child care squeeze: low-income parents worry and scramble to cope with county's shrinking safety net of subsidies. Democrat and Chronicle. February 7, 2004:1
Liang K-Y, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika. 1986;73 :13 –22
Hall CB, Douglas RG Jr. Respiratory syncytial virus and influenza. Practical community surveillance. Am J Dis Child. 1976;130 :615 –620
Hall CB, Walsh EE, Schnabel KC, et al. Occurrence of groups A and B of respiratory syncytial virus over 15 years: associated epidemiologic and clinical characteristics in hospitalized and ambulatory children. J Infect Dis. 1990;162 :1283 –1290
Long CE, Hall CB, Cunningham CK, et al. Influenza surveillance in community-dwelling elderly compared with children. Arch Fam Med. 1997;6 :459 –465
McConnochie KM, Roghmann KJ, Liptak GS. Hospitalization for lower respiratory tract illness in infants: variation in rates among counties in New York State and areas in Monroe County. J Pediatr. 1995;126 :220 –229
Diggle PJ, Heagerty P, Liang K-Y, Zeger SL. Analysis of Longitudinal Data. 2nd ed. Oxford, United Kingdom: Oxford University Press; 2002
American Academy of Pediatrics. 2003 Red Book: Report of the Committee on Infectious Diseases. 26th ed. Elk Grove Village, IL: American Academy of Pediatrics; 2003
American Public Health Association, American Academy of Pediatrics. Caring for Our Children. National Health and Safety Performance Standards: Guidelines for Out-of-Home Child Care Programs. Washington, DC: American Public Health Association; 2000
Pappas DE, Schwartz RH, Sheridan MJ, Hayden GF. Medical exclusion of sick children from child care centers: a plea for reconciliation. 2000;93 :575 –578
Landis SE, Earp JL, Sharp M. Day-care center exclusion of sick children: comparison of opinions of day-care staff, working mothers, and pediatricians. Pediatrics. 1988;81 :662 –667
Centers for Medicaid and Medicare Services. States where Medicaid reimbursement of services utilizing telemedicine is available. Available at: www.cms.hhs.gov/states/telelist.asp. Accessed March 12, 2005(Kenneth M. McConnochie, M)