Exploring the Acceptance of Telecare Among Senior Citizens: An Application of Back-Propagation Network

Abstract

Objective: This study employed a new method (nonlinear method) called back-propagation network [BPN] to analyze the adoption model of telecare. It is important to inform professionals of the need for medical statistics knowledge to better evaluate new technologies, particularly in health technologies assessment agencies. Methods: This study used face-to-face interviews to collect senior citizens aged over 60 years in Taiwan, and it used BPN based on the technology acceptance model to identify a telecare adoption model. Lastly, this study compared traditional methods and the BPN method. Results: The BPN method is much better than traditional method. Moreover, the main research result showed that people's perceived usefulness must be raised to effectively increase the adoption of telecare. Education and training for ease of use of telecare for users appear especially important. In sum, this research recommends making the operation interfaces of telecare more user-friendly and providing a demonstration system for practice, so that users would be more comfortable using the system. The findings of this research suggest that, to develop training for users to use telecare, it is likely to be helpful to reduce users' anxiety and improve usage of telecare. Conclusions: The utilization of the adoption model of telecare established by the BPN method of artificial neural networks is feasible. These findings may offer significant reference for subsequent studies.

Introduction

Telecare is the term used to describe the use of technology along with local clinical protocols to monitor remotely a patient's medical condition in their own home.¹ The advantages are that telecare provides cost benefits, improves the quality of life, and alleviates the problems of increasing medical expenses, demand for healthcare and accessibility to healthcare, and others.^2

–8 Hence, telecare will become a trend in the 21st century.^9,10

Telecare has been raised as a potential solution for many years without broad acceptance.¹¹ The equipment is perceived as not “user friendly” and bulky.¹² Further, evidence of benefit is unclear and there remains a need for robust studies before telecare can be recommended for widespread implementation. Koch and Hägglund¹³ also show that health informatics in elderly care is an expanding field of interest, but we still do lack knowledge about the elderly person's needs of technology and how it should best be designed.

On the other hand, most of previous studies used simple statistical methods or regression-based multivariate analysis, all of which are linear models. Few studies were carried out using nonlinear structural models (such as the nonlinear neural network model). Thus, the purpose of this study was to utilize the adoption model of telecare established by the back-propagation network (BPN) method. It is important to make professionals aware of the need of medical statistics knowledge to better evaluate new technologies, particularly in health technologies assessment agencies.

Technology Acceptance Model

Although research on technology acceptance has considerably developed in the area of information systems, the number of studies on technology acceptance model (TAM) in the field of healthcare, particularly in medical informatics, is rather limited. As a result, the development of TAM will be considered from an aspect of providing greater empirical healthcare.^6,14

–18

TAM^19,20 is primarily comprised of five constructs—perceived ease of use (PEOU), perceived usefulness (PU), attitude toward using (ATT), behavioral intention to use (BI), and actual system use. PU is defined as “the degree to which a person believes that using a particular system would enhance his or her job performance” and PEOU as “the degree to which a person believes that using a particular system would be free from effort.” Of these, PU and PEOU are the most important factors required for accepting new information technologies; further, PEOU affects PU and ATT affects BI directly.¹⁹

Telecare has been listed as a development design for newly emerging service industries for 2008 in Taiwan, although still in its infancy in Taiwan, and there are very few users who subscribe to this form of healthcare. As a result, BI rather than the actual usage is a better indicator of actual needs. Besides, Ajzen²¹ proposed the theory of planned behavior. He regarded that BI is a prerequisite for action and that only the ATT cannot directly predict behavior. In contrast, only BI enables accurate predictions.^21
–23 Hence, the research structure underlying this study is shown in Figure 1.

Fig. 1.

Research structure. PEOU, perceived ease of use; PU, perceived usefulness; ATT, attitude toward using; BI, behavioral intention to use.

Based on the theory of reasoned action by Fishbein and Ajzen,²⁴ it is believed that if the attitude of individuals toward behavior is more positive, their BI will be relatively stronger.^24,25 Further, assumptions 1–3 are obtained from relevant studies^19,20,24,26 based on TAM and theory of reasoned action.

H1: An individual's ATT and BI with respect to telecare are found to be positively associated.

H2a: PEOU has a direct effect on the ATT of telecare.

H2b: PEOU has a direct effect on PU.

H3: The stronger the PU of telecare, the stronger will be the ATT of telecare.

Back-Propagation Network

An artificial neural network (ANN) utilizes a large number of simple connected artificial neurons to imitate the capabilities of a biological neural network. It can be seen as a legitimate part of statistics that fits snugly in the niche between parametric and nonparametric methods.²⁷ ANN has been widely used to examine the complex relationship between input and output variables.²⁸ Of the various ANN models, the BPN is the simplest model and the easiest to understand. Hence, it is the most commonly used model at present.²⁹ The term “back propagation” (BP) indicates that the error correction signal/search direction goes backward from the output layer to the input layer.^30,31 The basic principle is that the sum of squared errors (the errors are the differences between the desired target values and the computed neural network outputs) is minimized by an iterative error correction mechanism that adjusts the weights between the neural network nodes until a stopping criterion is successfully reached.^30,32

Materials and Methods

Data Collection

This study used face-to-face interview to collect senior citizens aged over 60 years who lived in a village in Taiwan. A total of 295 valid copies of a questionnaire were obtained, with men accounting for 57% of the respondents. Most of them belonged to age groups of 60–69 (amounting to 58%). In terms of the educational level, 49% of the subjects had completed primary schools, and the average monthly income ranged between NT$20,000 and NT$50,000 (amounting to 71%) (US$1 ≈ NT$31.61).

Measures of the Constructs

This study is based on the definition as well as on the constructs related to TAM.^19,20 It includes four constructs: PEOU, PU, ATT, and BI. The operationalization and sources of measurement items in this study are shown in Table 1. All evaluation items employ a five-point Likert-type scale for measurement, wherein 1, 2, 3, 4, and 5 indicate “strongly disagree,” “disagree,” “fair,” “agree,” and “strongly agree,” respectively.

Table 1.

Reliability and Validity Results

Categories	Measure	Cronbach's Alpha
Perceived ease of use^19,20
PEOU1	I find that using telecare is simple.	0.81
PEOU2	I find that telecare is easy to learn.
PEOU3	I find that telecare is easily understandable and clear.
PEOU4	Overall, I find that using telecare is convenient.
Perceived usefulness and benefits^19,20
PU1	I find that using telecare is helpful in monitoring health.	0.59
PU2	Telecare can enhance my level of convenience in accessing medical care service.
PU3	Overall, I find telecare highly useful.
Attitude toward using³⁶
ATT1	I like using telecare.	0.66
ATT2	Overall, I consider telecare to be just right.
ATT3	In my old age, using telecare would be ideal.
Behavioral intention to use³⁶
BI1	Overall, I am highly willing to use telecare.	0.85
BI2	If necessary, I would use telecare often.
BI3	In my old age, I am willing to use telecare.
BI4	In my old age, I would use telecare often.

PEOU, perceived ease of use; PU, perceived usefulness; ATT, attitude toward using; BI, behavioral intention to use.

Reliability and Validity Analysis

The internal consistency was assessed using Cronbach's alpha (Table 1). The alpha coefficients (ranging from 0.6 to 0.9) of all the four constructs were higher than the benchmark of 0.6, as suggested by Bagozzi and Yi.³³ This demonstrated high internal consistency and, thereby, the reliability of each construct. As the content and viewpoints of the questionnaires for the facets of this study have referred to the viewpoints from related researches (refer to Table 1), the items of the questionnaires have included the facets of measurements. Therefore, the questionnaires of this study have considerable content validity.

Data Analysis Methods

Research framework of the ANN

The Feed Forward Neural Network has a general architecture,³⁴ as shown in Figure 2. The development procedure of ANN can be divided into two processes: learning process and recalling process. A large number of examples are required in both processes as the input data, for the ANN operation to complete at each process. During the learning process, the errors of trained examples with targets are used to adjust the weights of ANN by various efficient learning algorithms. The weights are updated after processing all examples. The final weights are adjusted to meet the minimization of errors we desired. Then, the trained weights are stored in the network after learning is completed.

Fig. 2.

Feed-forward neural network with n inputs and p outputs.

In the second process, testing examples are used to verify the validity of the network. Further, neural network is utilized in a case study to infer the results.

This study adopted three ANN frameworks that use a single hidden layer. They are shown in Figures 3 –5. BPN was used to adjust the network connection weights. The parameters of three ANNs are summarized in Tables 2 –4. In Figure 3, PEOU and PU are input layers, and ATT is the output layer. In Figure 4, PEOU is the input layer, and PU is the output layer. In Figure 5, ATT is the input layer, and BI is the output layer.

Table 2.

Parameters for Artificial Neural Network Research Framework Diagram (A)

	Input		Hidden		Output
Number of neurons	2		2		1
Signal function	Linear		Sigmoidal		Linear
Neuron index range	i = 0,1		h = 0,1		j = 0
Weights (including bias)		→ Wih →		→ Whj →

Table 3.

Parameters for Artificial Neural Network Research Framework Diagram (B)

	Input		Hidden		Output
Number of neurons	1		1		1
Signal function	Linear		Sigmoidal		Linear
Neuron index range	i = 0		h = 0		j = 0
Weights (including bias)		→ Wih →		→ Whj →

Table 4.

Parameters for Artificial Neural Network Research Framework Diagram (C)

	Input		Hidden		Output
Number of neurons	1		1		1
Signal function	Linear		Sigmoidal		Linear
Neuron index range	i = 0		h = 0		j = 0
Weights (including bias)		→ Wih →		→ Whj →

Fig. 3.

ANN research framework diagram (A). ANN, artificial neural network.

Fig. 4.

ANN research framework diagram (B).

Fig. 5.

ANN research framework diagram (C).

Data analysis steps

In this study, we constructed, performed training on, and tested the ANN model with Matlab Neural Network Toolbox.

Step 1: Normalize all data into [0, 1], including PEOU and PU.

Step 2: The trained algorithm of the Levenberg-Marquardt backpropagation³⁵ is used to adjust the weight with mean square error.

Step 3: Randomly choose the initial weights.

Step 4: Proceed with the ANN model training; 2/3 of the total samples (197 samples) are chosen as training examples. Each training session for all 197 data is deemed as an epoch. Parameters are updated during this period to obtain the optimal ANN models.

Step 5: The final optimal ANN model is acquired after completing the training. Tests (a) and (b) are employed to examine the stability and reliability of ANN; tests (c) and (d) are employed to estimate the effect of individual input variables on output variables.

(a) Input the 197 samples, which have been used in training, to obtain the 197 estimated output values. The errors between the estimated and raw data are shown in Figure 6. The statistical result of errors, including mean, standard deviation (SD), and root of mean square (RMS), is shown in Table 5.

(b) Input the remaining 1/3 of the samples (98 samples), which have not been used in training, to obtain the 98 estimated output values. The errors between the estimated and raw data are shown in Figure 6. The statistical result of errors, including mean, SD, and RMS, is shown in Table 5.

(c) Take the dependence of ATT on PEOU as an example; the correlation with ATT could be obtained by setting PEOU to 1 and other input variables to zero. Thus, the test vector of input is [1 0 0 0 0]. Similarly, the correlation between all input variables and output variables could be acquired (Table 5).

(d) Besides test (c), there is another approach to identify the effect of one input variable on output variables, and it is performed by fixing other input variables to zero. Thus, continuous discrete input is used with one input variable increasing from 0 to 1 at 0.01 steps. Similarly, the correlation between all input and output variables could be acquired. The results are presented in Figures 7 –9. Therefore, the dependence of all output variables on input variables can be identified.

Results

Analysis Results of the Artificial Neural Network Model

Table 5 shows the dependence of ATT on PEOU and PU. The error mean (SD) and RMS for the training examples are 0.1415 (0.0790) and 0.1620, respectively; the error mean (SD) and RMS for the testing examples are 0.1439 (0.0806) and 0.1645, respectively. As to the dependence of PEOU on PU, the error mean (SD) and RMS for the training examples are 0.1263 (0.0879) and 0.1538, respectively; the error mean (SD) and RMS for the testing examples are 0.1192 (0.0850) and 0.1459, respectively. As to the dependence of ATT on BI, the error mean (SD) and RMS for the training examples are 0.1035 (0.0962) and 0.1411, respectively; the error mean (SD) and RMS for the testing examples are 0.0475 (0.0438) and 0.0643, respectively. As shown, the errors are acceptable. Further, it could also be inferred from Figure 6 that the errors for the training and testing examples in the ANN research framework diagrams (A), (B), and (C) are acceptable. Also, it means that the model of ANN constructed in this study is stable and reliable.

Fig. 6.

Errors of the training and testing examples in the ANN research framework diagrams (A), (B), and (C). (a) Errors of the training examples in the ANN research framework diagram (A). (b) Errors of the testing examples in the ANN research framework diagram (A). (c) Errors of the training examples in the ANN research framework diagram (B). (d) Errors of the testing examples in the ANN research framework diagram (B). (e) Errors of the training examples in the ANN research framework diagram (C). (f) Errors of the testing examples in the ANN research framework diagram (C).

As shown in Figure 7, ATT varies positively with BI and thus H1 is supported. As shown in Figure 8, PEOU varies positively with PU and thus H2b is supported. As shown in Figure 9, PEOU and PU have a significantly positive effect on ATT, and thus H2a and H3 are supported. Besides, it also reveals that the impact of PU on ATT with respect to telecare is higher than that of PEOU.

Fig. 7.

Effects of ATT on BI.

Fig. 8.

Effects of PEOU on PU.

Fig. 9.

Effects of PEOU and PU on ATT. (a) Effects of PEOU on ATT; (b) effects of PU on ATT.

In addition, the correlation coefficients in Table 5 show that, among the effects of PEOU and PU on ATT, the effect of PU is the most significant (with a correlation coefficient of 0.4903) and positive, followed by PEOU (with a correlation coefficient of 0.1639), which is also positive. The results of above three hypotheses testing indicate that it is feasible to construct a model with BPN method of ANN.

Table 5.

Analysis Results of the Artificial Neural Network Model

		Train Error		Test Error
Path	Correlation	Mean (SD)	RMS	Mean (SD)	RMS
PEOU → ATT	0.1639	0.1415 (0.0790)	0.1620	0.1439 (0.0806)	0.1645
PU → ATT	0.4903	0.1415 (0.0790)	0.1620	0.1439 (0.0806)	0.1645
PEOU → PU	0.9817	0.1263 (0.0879)	0.1538	0.1192 (0.0850)	0.1459
ATT → BI	0.9722	0.1035 (0.0962)	0.1411	0.0475 (0.0438)	0.0643

SD, standard deviation; RMS, root of mean square; BI, behavioral intention to use.

Results of Traditional Method and the Artificial Neural Network Model

Table 6 reveals the dependence of ATT on PEOU and PU. The training and testing data RMSs are 0.162 and 0.165, respectively. As to the dependence of PEOU on PU, the training and testing data RMSs are 0.154 and 0.146, respectively. As to the dependence of ATT on BI, the training and testing data RMSs are 0.141 and 0.064, respectively. On the other hand, the regression model also shows the dependence of ATT on PEOU and PU. The training and testing data RMSs are 0.809 and 0.557, respectively. As to the dependence of PEOU on PU, the training and testing data RMSs are 0.577 and 0.406, respectively. As to the dependence of ATT on BI, the training and testing data RMSs are 0.748 and 0.297, respectively. In brief, this comparison clearly shows that the ANN model is much better than the regression method.

Table 6.

Comparison Between Traditional Method and the Artificial Neural Network Model

Model	Path	Training Data RMS	Testing Data RMS
ANN	PEOU → ATT	0.162	0.165
	PU → ATT	0.162	0.165
	PEOU → PU	0.154	0.146
	ATT → BI	0.141	0.064
Regression	PEOU → ATT	0.809	0.557
	PU → ATT	0.809	0.557
	PEOU → PU	0.577	0.406
	ATT → BI	0.748	0.297

ANN, Artificial Neural Network.

Discussion and Conclusions

This research has proved that the utilization of the adoption model of telecare established by BPN is feasible. The results identify telecare adoption model by using BPN based on TAM. Using the BPN model, we do not need to be restricted in the research data whether for linear or nonlinear.

In addition, the TAM reveals that ATT and BI with respect to telecare is positively and significantly associated, and this result is consistent with those of relevant past studies.^17,24,25 People's PU must be raised to effectively increase the adoption of telecare. PU has significant positive effect on users' attitude, indicating the attitude of use will be positively affected when users find the system to be helpful, that is, the users will adopt telecare if they perceive the system useful. This result is consistent with that obtained from the investigation on physician's acceptance of telemedicine technology—PU was a critical determinant of attitude—conducted by Chau and Hu¹⁷ as well as in previous application researches^{16,37

–41} on TAM.

Further, education and training for ease of use of telecare for users appear especially important. This result is similar to the study by Ludwick and Doucette,⁴² which indicated that the telemedicine system uses ordinary videoconferencing, whose lack of complexity helps reduce telemedicine barriers. The result of this research is consistent with those obtained from other researches^{14,19,20,26,37,38,40} that used TAM. Telecare can provide cost benefits and improve the quality of life.^43,44 In sum, this research is suggested to make the operation interfaces of the telecare more user-friendly and provide a demonstration system for practice, so that users would be more comfortable using the system. The findings of this research suggest that, to develop training for users to use telecare, it is likely to be helpful to reduce users' anxiety and improve usage of telecare. Relevant studies also reveal similar results and indicate that training should be provided to professionals.^45,46 Participants of health professionals had high levels of satisfaction regarding knowledge and skills, use of information to enhance patient care, program quality, and convenience of the technology.⁴⁷

This research is different from the general researches that focus merely on TAM for singular analysis because it can help enhance the understanding and reference pertaining to the relationship and reference of important user variables in the choice of telecare by relevant researchers, technology developers, and policy makers. These findings may offer significant reference for subsequent studies.

Disclosure Statement

No competing financial interests exist.

References

Smith

, Elkin

, Partridge

. Technology and its role in respiratory care. Prim Care Respir J, 2009; 18:159–164.

Buysse

, Moor

, Maele

, Baert

, Thienpont

, Temmerman

. Cost-effectiveness of telemonitoring for high-risk pregnant women. Int J Med Inform, 2008; 77:470–476.

Darkins

, Ryan

, Kobb

, Foster

, Edmonson

, Wakefield

et al. Care coordination/home telehealth: The systematic implementation of health informatics, home telehealth, and disease management to support the care of veteran patients with chronic conditions. Telemed E Health, 2008; 14:1118–1126.

Rojas

, Gagnon

. A systematic review of the key indicators for assessing telehomecare cost- effectiveness. Telemed E Health, 2008; 14:896–904.

Prinz

, Cramer

, Englund

. Telehealth: A policy analysis for quality, impact on patient outcomes, and political feasibility. Nurs Outlook, 2008; 56:152–158.

Rajasekaran

, Radhakrishnan

, Subbaraj

. Elderly patient monitoring system using a wireless sensor network. Telemed E Health, 2009; 15:73–79.

Topacan

, Basoglu

, Daim

. Exploring the success factors of health information service adoption Proceedings of the PICMET: Portland International Center for Management of Engineering and Technologyart. no. 45998712008; 2453–2461.

Vimarlund

, Olve

, Scandurra

, Koch

. Organizational effects of information and communication technology (ICT) in elderly homecare: A case study. Health Inform J, 2008; 14:195–210.

Haux

. Health information systems—past, present, future. Int J Med Inform, 2006; 75:268–281.

10.

Miller

. Solving the disjuncture between research and practice: Telehealth trends in the 21st century. Health Policy, 2007; 82:133–141.

11.

Evers

, Loey

. Monitoring patient/ventilator interactions: Manufacturer's perspective. Open Respir Med J, 2009; 12:17–26.

12.

Horton

. The use of telecare for people with chronic obstructive pulmonary disease: Implications for management. J Nurs Manage, 2008; 16:173–180.

13.

Koch

, Hägglund

. Health informatics and the delivery of care to older people. Maturitas, 2009; 63:195–199.

14.

Kim

, Chang

. Key functional characteristics in designing and operating health information websites for user satisfaction: An application of the extended technology acceptance model. Int J Med Inform, 2007; 76:790–800.

15.

Schaper

, Pervan

. ICT and OTs: A model of information and communication technology acceptance and utilisation by occupational therapists. Int J Med Inform, 2007; 76S:S212–S221.

16.

Zheng

, Padman

, Johnson

, Diamond

. Evaluation of healthcare it applications: The user acceptance perspective. Stud Comput Intell, 2007; 65:49–78.

17.

Chau

PYK

, Hu

PJH

. Investigating healthcare professionals' decisions to accept telemedicine technology: An empirical test of competing theories. Inf Manage, 2002; 39:297–311.

18.

Raghupathi

. Health care information systems. Commun ACM, 1997; 40:80–82.

19.

Davis

. Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Q, 1989; 13:319–339.

20.

Davis

, Bagozzi

, Warshaw

. User acceptance of computer technology: A comparison of two theoretical models. Manage Sci, 1989; 35:982–1003.

21.

Ajzen

. From intentions to actions: A theory of planned behavior, action-control: From congition to behavior. Heidelberg: Springer, 1985.

22.

Ajzen

. Attitude, personality, and behavior. Milton Keynes: Open University Press, 1989.

23.

Ajzen

. The theory of planned behavior, organizational behavior and human. Decis Processes, 1991; 50:179–211.

24.

Fishbein

, Ajzen

. Belief, attitude, intention and behavior: An introduction to theory and research. Reading, MA: Addison-Wesley, 1975.

25.

Ajzen

, Fishbein

. Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall, 1980.

26.

Legris

, Ingham

, Collerette

. Why do people use information technology? A critical review of the technology acceptance model. Inf Manage, 2003; 40:191–204.

27.

Aldrich

. Exploratory analysis of metallurgical process data with neural networks and related methods. South Africa: Elsevier, 2002; 5.

28.

Nelson

, Illingworth

. Practical guide to neural nets. Addison Wesley Publishing Company, 1994.

29.

Davies

, Goode

, Mazanec

, Moutinho

. LISREL and neural networks modeling: Two comparison studies. J Retailing Consum Serv, 1999; 6:249–261.

30.

Chen

, Jain

. A robust back propagation learning algorithm for function approximation. IEEE Trans Neural Netw, 1994; 5:467–479.

31.

Werbos

. The roots of backpropagation. New York: Wiley and Sons, Inc., 1994.

32.

Bagozzi

, Yi

. On the evaluation of structural equation models. J Acad Market Sci, 1988; 16:74–94.

33.

Bishop

. Neural networks for pattern recognition. Oxford University Press: New York, 1995.

34.

Kumar

. Neural networks: A classroom approach, 4th. New Delhi: Tata McGraw-Hill, 2007.

35.

Hagam

, Menhaj

. Training feedforward networks with the Marquardt algorithm. IEEE Trans Neural Netw, 1994; 5:989–993.

36.

Taylor

, Todd

. Understanding information technology usage: A test of competing models. Inf Syst Res, 1995; 6:144–176.

37.

Venkatesh

, Morris

, Davis

. User Acceptance of information technology: toward a unified view. MIS Q, 2003; 27:425–478.

38.

Agarwal

, Prasad

. Are individual differences germane to the acceptance of new information technologies? Decis Sci, 1999; 30:361–391.

39.

Keil

, Beranek

, Konsynski

. Usefulness and ease of use: Field study evidence regarding task considerations. Decis Support Syst, 1995; 13:75–91.

40.

Schepers

, Wetzels

. A meta-analysis of the technology acceptance model: Investigating subjective norm and moderation effects. Inf Manage, 2007; 44:90–103.

41.

Hung

, Chang

. User acceptance of WAP services: Test of competing theories. Comput Stand Interfaces, 2005; 27:359–370.

42.

Ludwick

, Doucette

. Primary care physicians' experience with electronic medical records: Barriers to implementation in a fee-for-service environment. Int J Telemed Appl, 2009[Epub ahead of print]doi:10.1155/2009 = 853524.

43.

Ferrer-Roca

, Garcia-Nogales

, Pelaez

. Telemed E Health, 2010; 16:233–243.

44.

Michelle

, Tom

, Patty

, Wayne

, Michael

, Matthew

, Manya

. Provider satisfaction and patient outcomes associated with a statewide prison telemedicine program in Louisiana. Telemed E Health, 2010; 16:472–479.

45.

Albert

, Justin

, David

, Eneida

, Walter

, Jason

, Steven

. IDEATel Consortium. The Remote Patient Education in a Telemedicine Environment Architecture (REPETE) Telemed E Health, 2008; 14:355–361.

46.

David

, Shari

, Sue

, Kimberly

, Karen

, Charles

. Incorporating telehealth into primary care resident outpatient training. Telemed E Health, 2009; 15:277–282.

47.

Ann

, Cathy

, Betty

. Satisfaction with a distance continuing education program for health professionals. Telemed E Health, 2010; 16:776–786.