Thirty two studies met our selection criteria. The studies examined the use of interactive video in diverse contexts ranging from specialist consultations to home nursing. Many of these represented demonstration and feasibility studies rather than full scale trials. This is reflected in their sample sizes often being small and selection criteria for study participants rarely being random in nature. Only seven studies had more than 100 participants,5–11
14 were small pilot studies with less than 100 patients,12–25
and 10 were simple feasibility studies with 20 or fewer patients.26–35
One paper, which presented an overview of an Australian regional telepsychiatry project, did not provide patient numbers.36
The table lists the studies by type of consultation. (An extra table on the BMJ 's website provides further detail of studies in which patient numbers were over 20 and methods of measuring patient satisfaction were explicitly described. None of these studies declared any conflicts of interest.)
In terms of methodologies used, 26 studies used simple survey instruments, five did not specify the exact methods, and one used qualitative methods. Only one study was a randomised controlled trial,10
in two others patients were randomly selected,19,23
and one was a case-control study.20
In the remaining 28 studies selection criteria were not specified or participants represented consecutive referrals, convenience samples, or volunteers.
Measures of patient satisfaction
The studies mainly used simple survey instruments to ascertain patient satisfaction. Firm conclusions are limited by methodological difficulties, but it would seem that the patients found teleconsultations acceptable; noted definite advantages, particularly increased accessibility of specialist expertise, less travel required, and reduced waiting times; but also had some disquiet about this mode of healthcare delivery, particularly relating to communication between provider and client via this medium.
Shortcomings of studies
We identified several problems with the studies that affect their reliability and validity. Many studies had small sample sizes, almost a third having 20 or fewer participants, and low response rates, as low as 50%.22
Patient selection criteria were often not clearly specified, or there were no formal selection criteria. Most of the studies (28) used volunteers or physician referrals and provided no information about refusal rates at point of initial referral. Thus, it is not possible to discount selection bias in favour of those likely to be positive about teleconsultation.
Methodologies used for assessing satisfaction were not clearly specified in many studies, making interpretation and comparison of results problematic. Most studies sought to measure whether patients would use the systems again or were “satisfied” with the service. However, few studies defined what satisfaction meant. Therefore, we are unable to discern whether the participants said they were satisfied because telemedicine didn't kill them, or that it was “OK,” or that it was a wonderful experience. The available evidence does not help us to understand the reasons underlying satisfaction or dissatisfaction. In addition, most of the studies presented only initial impressions and failed to explore what happened to patient satisfaction over time, thereby making it possible that the novelty value of the technology resulted in a positive bias.
The cost of teleconsultations compared with routine consultations was not addressed. This is particularly pertinent to the US studies, which account for over 45% of the studies found. The US system of healthcare delivery is a fee for service system, yet the US studies do not mention whether patients attending for teleconsultation paid for the service in the usual way or whether they received this service free of charge. As many US telemedicine projects are primarily grant funded, it is possible that in some studies participants received free teleconsultations, which could affect their satisfaction with the service provided.
Because of the survey nature of most of the studies, there are often inconsistencies in responses that remain unexplained. One possible explanation lies in the survey design. Many surveys have questions with multiple constructs (such as: “I felt the physician was easy to talk to and understood everything I said”). When a single question contains two constructs it is not possible to know which actual construct the participant is responding to, making the data difficult to interpret.
The effects, if any, of telemedicine on communicative behaviours and the interaction between provider and patient during the consultation remained virtually unexplored. There was a lack of data examining patients' perceptions.
Generalisability of results
The generalisability of much of the published research is limited because of effect modifiers such as study setting. One of the largest studies examined teleconsultation in a prison in the United States.5
Clearly, there are several reasons why satisfaction in prisoners may be different from that in the general population. Thus, the peculiarities of the setting mean that this study's results cannot be applied reliably to the general population of that country or more widely.
Furthermore, the delivery of health care was somewhat artificial in many studies. Participants often received a teleconsultation in addition to a routine consultation, and so were really being asked to make a hypothetical judgment as to its value. In many studies participants also received “special” treatment, with every effort being made to minimise inconvenience. Satisfaction in these somewhat artificial contexts may not be readily translatable to satisfaction with telemedicine when it is being used in routine practice.