Abstract
We examined the feasibility of using home blood pressure (BP) telemonitoring devices for managing patients with poor BP control. We enrolled 591 subjects with a diagnosis of hypertension. Patients were randomized to usual care (n = 147) or to the intervention arm (n = 441). Those in the intervention arm were issued with a home BP telemonitoring device. The device transmitted BP readings automatically via the home telephone line. Technical alerts were generated if patients did not transmit their BP readings according to the protocol. During the first six months, 693 technical alerts were generated by 267 patients. About half of these patients (112) generated more than two technical alerts. Resolution of the alerts showed that 61% were caused by patient non-adherence. Patients who generated >2 technical alerts were younger (61 vs. 64 years; P = 0.001) and were more likely to be non-Caucasian (64% vs. 47%, P = 0.002) than those generating 2 or fewer alerts. Despite the potential for improving health care using home BP telemonitoring, certain patients will require more support to use the equipment successfully.
Introduction
A growing body of evidence suggests that when done properly, home blood pressure (BP) readings are more prognostically significant than those taken at the clinic. 1 In addition, managing hypertension based on home BP readings may result in decreased medication and overall cost in comparison to using clinic measurements. 2,3 One way to improve BP control is by BP telemonitoring in patients' homes.
Despite the fact that health insurers are now covering certain telemedicine services, the feasibility of providing many of these services has not been well studied. 4 Although there is growing enthusiasm for BP telemonitoring, it is not clear how feasible this is for those who may benefit the most. In addition, it is not clear that all individuals are capable or well-suited for using home BP telemonitoring. We have investigated the feasibility of home telemonitoring among primary care patients with poor blood pressure control.
Methods
The Hypertension Intervention Nurse Telemedicine Study (HINTS) 5 is a randomized clinical trial. The present paper provides information about the first six months of experience in using the BP monitoring equipment.
Study population
A total of 8221 veterans with a diagnosis of hypertension were identified from the Durham VA Medical Center. The diagnosis was based on ICD-9 outpatient codes for hypertension, at least one primary care visit in the previous year, and an average BP of >140/90 mm Hg during the previous year. The patients were not required to have a computer or be computer literate in order to participate in the study. Neither were the patients required to have prior knowledge of how to use the blood pressure monitor. A total of 1542 patients met the preliminary inclusion/exclusion criteria and were sent letters of invitation. We enrolled 591 patients, but three patients were later excluded. Of the 588 included in the present study, we randomized 147 patients to usual care and 441 patients to receive one of three interventions. Only patients randomized to the intervention groups were issued BP telemedicine equipment. The study was approved by the appropriate ethics committee.
The intervention arms were: (1) a nurse-administered, tailored behavioural intervention; (2) a nurse-administered medication management according to a hypertension decision support system; and (3) a combination of these two interventions. The interventions were triggered based on home BP values transmitted via telemonitoring devices over ordinary telephone lines to a server.
Home BP training
Each patient randomized to the intervention was provided with user instructions at the initial recruitment visit. The research assistant used this material to train patients in the use of the equipment and gave instructions to take three BP readings a week, on different days. Patients were also informed that if we did not receive readings for a two-week period they would be contacted and asked to verify the status of the equipment. The information pack also instructed patients on what to do when they obtained high readings (systolic BP >180 mmHg), or needed to contact study staff, or had questions about the equipment. The information pack gave patients step-by-step instructions with pictures of how to set up the equipment in their homes. The information included instructions about what to do if participants were going to be away from their home for longer than one week. Instructions for resetting the equipment after power failures were also included in the information pack.
Monitoring equipment
A blood pressure monitor (A&D 767PC, CareMatix Inc) was used which could transmit data via a wireless link to a home hub. The hub was connected to the home telephone line. The cost of the monitoring equipment was $550 per device. We chose the wireless home BP monitor and telemedicine hub because we thought it would be more convenient for patients, which might increase adherence to taking BP readings. The patient could set up the hub anywhere near a telephone socket and electrical outlet in their home.
When patients measured their BP and pulse, readings were sent to the server automatically. The BP and pulse readings were accessible to the study team via a website. An automated interface processed the readings and updated the patient record with the BP and pulse reading every hour. This information was used to activate the intervention if necessary.
Alert manager
The alerting process was automated. An algorithm reviewed the transmitted readings and triggered a technical alert if there were too few readings to make a determination of adequate BP control. The technical alerts were generated when the patient had not transmitted three or more readings on three or more different days during a two-week period. Patients were asked to take six readings – one every other day over two weeks. Technical alerts were classified as either system technical or patient initiated. A system technical alert meant that the hypertension intervention algorithm noted that there were insufficient BP readings to identify whether the patient had adequate BP control. A patient-initiated alert resulted from a patients' report that there appeared to be a problem with the device.
Alert resolution
Resolution of the alerts was conducted by the nurses after they had contacted the patients and provided the necessary intervention materials. The technical alerts were classified into one of five categories based on information obtained by the research assistant:
Patient knowledge, if the research assistant decided that the patient had not fully understood what was needed to take and transmit BP readings; Patient non-adherence, if (a) the patient responded to a message from the research assistant mentioning that no readings had been transmitted and subsequently transmitted at least a single BP reading or, (b) the patient transmitted only a single BP reading after the alert creation date following instructions from the research assistant to carry out a system reset, or (c) the patient admitted to not following the study protocol; Device problem, if the patient indicated that they were unable to take readings because the devices were not working correctly; System problem, if the patient indicated that they had been taking readings, yet no data had been received and after resetting device, readings were received which included dates prior to the alert date; Other, if the patient had been hospitalized, notification of vacation plans had not been received, or the patient was in the process of receiving replacement devices.
On average, it took the research assistant four days to resolve each technical alert.
Analysis
We divided the data according to the number of alerts: 0–2 vs. >2. We calculated the percentage of intervention subjects with 0–2 vs. >2 technical alerts for baseline demographic characteristics that might be associated with an increased likelihood of triggering technical alerts and also calculated means and SDs for continuous variables of interest. We tested for differences in these baseline characteristics between intervention subjects using a two-tailed chi-square test for categorical variables and t-tests for continuous variables. We collapsed race and employment status into 2 level variables (Caucasian vs. non-Caucasian; retired vs. not retired) for the chi-square tests.
Results
Of the 588 participants included in the study, 441 were randomized to receive home BP telemonitoring: (1) nurse-administered, tailored behavioural intervention (n = 148); (2) nurse-administered medication management according to a hypertension decision support system (n = 147); (3) combination of the two interventions (n = 146), while 147 were randomized to usual care.
Technical alerts
During the 6-month study, a total of 693 technical alerts were generated by 267 of the intervention patients (174 patients had no technical alerts). The demographic characteristics of the 441 intervention subjects are summarised in Table 1, stratified by the number of technical alerts (0–2 vs. >2) that were generated. Nearly 75% (329/441) of intervention patients triggered two or fewer technical alerts during that time; 112 patients (25%) triggered more than two technical alerts. Subjects who triggered more than two technical alerts were younger (60.5 years vs. 64.5; P = 0.0006), less likely to be Caucasian (36% vs. 53%, P = 0.002) and less likely to be retired (38% vs. 54%, P = 0.003) than were subjects who triggered two or fewer technical alerts. In addition, those who triggered two or fewer technical alerts adhered more closely to the study protocol by transmitting an average of 5.8 (SD = 2.8) BP readings over two-week intervals than did those who triggered more than two technical alerts with an average of 3.1 (SD = 1.1) (P < 0.0001).
Characteristics of the 441 intervention subjects, stratified by the number of technical alerts generated
aAll data except gender and diabetic status were patient-reported. When missing values existed, they were included in the calculation of percentages
bThe chi-square P values for categorical variables and the P values from the t-test for continuous variables are for the differences between 0–2 vs. >2 technical alerts
cThe chi-square P value was calculated for Caucasian vs. non-Caucasian categorization
dThe chi-square P value was calculated using retired vs. all other employment categories
eInadequate income was defined as reporting difficulty in paying bills
fHealth literacy <9th Grade corresponds to a REALM 10 score ≤60
gThe average number of BP readings transmitted was the average for the 13 patient specific determination periods during the 6 months of the study
Alert resolution
Four hundred and twenty-one (61%) of the technical alerts were attributed to patient non-adherence (Table 2). A small proportion of the technical alerts (5%) were attributed to a lack of patient knowledge. These were due to the patient having difficulty in setting up equipment, putting on the BP cuff and/or understanding how many readings to take per week.
Number and type of technical alerts occurring in six months
Eighty-nine (13%) of the technical alerts were due to device problems. We replaced the equipment when we were unable to get the devices to transmit from the participants' homes. After subsequent testing, we sent 31 BP telemonitoring devices back to the vendor for repair.
System problems were responsible for 54 (8%) of the technical alerts. A review identified that telephone disconnections, loss of power or inconsistent telephone service were the main causes. Once the research assistant had instructed the patient to carry out a device reset, we often received data that had been previously stored in the home equipment.
Ninety-four (14%) of the technical alerts were classified as ‘other’. Some of these alerts were due to patients being hospitalized for emergency reasons and the study team not finding out about the hospitalization until after the alert had triggered. In some cases the alerts involved patients being in the process of receiving replacement devices and thus not having sufficient time to transmit readings. Other cases were due to communication lapses between study team members resulting in a patient's status not being correctly updated.
Discussion
During the six-month study, 75% of the intervention patients were able to set up the devices and adhere to the study protocol. However, one-quarter of them were unable to adhere to the study protocol. It was discouraging that 61% of the technical alerts were due to patient non-adherence. Non-adherence represents a significant barrier for the use of telemedicine interventions. Patients who triggered more than two technical alerts were only transmitting about three BP measurements every two weeks, whereas those participants who triggered two or fewer technical alerts transmitted nearly twice as many. The patients who triggered two or fewer technical alerts adhered more closely to the described study protocol by either taking readings as requested or notifying the study team when readings were not to be expected.
Employment status, age and race were the few demographic characteristics that appeared to be associated with the number of technical alerts triggered by subjects. Older subjects and subjects who were retired were more likely to use the BP monitoring system according to the guidelines. Conversely, minority subjects were less likely to follow the study guidelines in terms of BP reading transmission or notification to the study team to not expect readings.
We would have expected older participants to have had more difficulty than younger subjects with using the monitoring equipment. However, the four-year difference in the mean ages of the groups who were adherent versus non-adherent is, in practical terms, small. A possible reason that older adults were more adherent to the study protocol could be that they had more comorbidities and contact with the health-care system. Support for this explanation comes from Ralston et al. 6 who observed that patients with more comorbid conditions were more likely to use a patient web portal with web communications than younger adults. Consistent with other areas, minorities, in this case African-Americans, may be less adherent to the study protocol because of competing needs as a result of more comorbidites. 7,8 The remaining demographic factors that we examined did not suggest any particular difference between those who were adherent and those who were non-adherent to the study protocol.
Two research assistants were initially employed full-time on the project. During the first four months of the study, each would spend 1–2 hours a day handling the technical alerts. Prior to the first six-month follow-up visit, we increased the staffing by another full-time research assistant. This was done in order to have a dedicated person available to handle technical alerts and facilitate the follow-up visits efficiently. The research assistants spent approximately 10 minutes training patients to take their own blood pressure and demonstrating how to set up the equipment in their homes. A number of the patients when contacted about the lack of readings transmitted commented that ‘they had just taken it the day before’, when actually a few days had passed. When this was explained to them they indicated that it was not yet part of their routine. Pickering et al. 9 noted that the American Heart Association recommends that patients take two or three home BP readings each morning and night, thus providing up to six daily readings for clinical decision making. In hindsight, perhaps our decision not to overburden the patients with taking readings every day may have had the unintentional consequence of not providing enough structure in the protocol for some of our patients, thus making it difficult for participants to remember whether they had taken a reading or not the previous day.
Device or system alerts accounted for 143 (21%) of the technical alerts. A few of the device problems involved the failure of the blood pressure monitor to connect to the hub, the monitor not correctly inflating or the monitor giving error messages instead of readings. In these cases we replaced the devices. Most of the problems were solved by asking the patient to reset the device, which involved disconnecting the hub from the power and telephone sockets and then plugging them back in. The system and device technical problems remain a cause for concern, both because of the frequency with which losses of power and telephone service occurred for many patients in smaller towns and rural areas, and also because of the additional staffing resources required to resolve the consequent alerts.
An additional barrier to successful transmission of BP readings is that the telemonitoring hub did not work seamlessly with all types of telephone service. As patients migrated from one telephone service to another, or had different telecommunications equipment installed at their homes, re-initialization of the hub was required. On some occasions these telecommunication changes prevented the hub from working. The inability of the hub to re-establish a connection when it lost power caused patients to question whether they were following procedures correctly. It can be expected that as the monitoring technology becomes more popular, many of the technical problems will be solved as the equipment becomes more standardized, in much the same way as have personal computers and mobile phones.
Patient knowledge resolutions accounted for only 35 (5%) of the alerts generated. In all of these cases, on contacting the patients, it was found that they had misunderstood an important piece of information during the initial training. Most of these alerts were resolved relatively quickly. In three instances, the participant informed the research assistant that this was too difficult for them to do and they could not participate in this part of the study.
The study participants came from a pool of veterans who received their primary health care through the Durham VA primary care clinics. This may limit the generalizability of the findings to other settings.
Conclusion
BP telemonitoring equipment is becoming more readily available and most patients are able to use BP monitors and understand the value of the device. Is it feasible to use BP telemonitoring among primary care patients in the VA? While our study was successful in getting 75% of the patients to use the devices correctly, there remain barriers to the seamless use of this technology in the primary care clinics. There are still some patients for whom this technology may not be appropriate or is problematic; patients may be either unwilling (non-adherent) or unable to fulfil the intervention protocol. Investigators and clinicians using these devices therefore need to pay attention to the population they are enrolling and allocate the necessary time to engage the patients whose work schedules may interfere with their involvement. In addition, the study team should be prepared for changes in the telephone system and device infrastructure, and have alternative data collection methods available. Such methods could include using postal transmission of data, or use of technology not dependent on the patients' telephone service such as mobile phones or satellite communication. The increased time and attention required by the non-adherent patients or those unable to manage the technology may have significant cost implications. Further evaluation of BP telemonitoring is required before the technique is widely implemented in health care.
Footnotes
Acknowledgements
This research was supported by a grant from Veterans Affairs, Health Services Research and Development (IIR 04-426), a VA Senior Career Scientist Award, and an Established Investigator Award from American Heart Association to the last author.