Use of short message service (SMS) reminders to improve the attendance rate of diabetes-related multidisciplinary services in a primary care clinic a randomized controlled trial

Background: of This study determined whether using reminders by Short Message Service (SMS) in patients with type II diabetes attending a primary care clinic resulted in higher attendance rates of NPLM services when compared with those without SMS reminders. Method: A single centre randomized-controlled trial included adult subjects with type II diabetes who could read Chinese or English, possessed an SMS-capable phone and booked at least NPLM service appointments in study site. The main outcomes were comparing attendance rates between the control group and the intervention group of various NPLM services including nurse complication screening, nurse individual counselling, multidisciplinary group education, dietitian counselling, physiotherapist service, retinopathy screening by optometrist and blood taking sessions Results: 297 eligible subjects were approached, 15 were excluded and 11 refused to participate. 271 subjects were recruited (response rate 96%). Their mean (± standard deviation) age was 63.3±11.3. Their mean duration of DM was 6.9±5.3 years with mean HbA1c 7.37±1.32%. They were randomly assigned to intervention (N=135) or control (N=136) group. The Intervention groups received a standard SMS reminder included appointment name, date and time 24-72 hours prior to the scheduled appointment. The Control groups received no reminder. 733 appointments (intervention N=372, control N=361) were analysed. More subjects used smartphone (82.2%) than mobile phone without internet access (17.8%). All SMS were successfully delivered for the 372 appointments with no adverse event reported. The overall attendance rates were 92.7% in the intervention group and 83.1% in the control group (Odds ratio, OR 2.60, 95% confidence interval, CI 1.61-4.19, p<0.001). In subgroup analysis, SMS reminders were particularly useful in reminding nurse clinic (OR 3.06, 95% CI 1.22-7.66, p=0.017) and optometrist clinic (OR 2.73, 95% CI 1.07-7.01,

3 Evidence had shown that multidisciplinary care can improve the quality of care including process and clinical outcomes of patients with type 2 diabetes mellitus (DM) (1,2). The American Diabetes Association recommended multidisciplinary care for patients with DM: such as participation in diabetes self-management education; receiving medical nutrition therapy, preferably by a registered dietitian; engagement in physical activity and regular eye examination by ophthalmologist or optometrist (3,4).
However, the beneficial effects would be negated if patients failed to attend those non-physician-led multidisciplinary (NPLM) services. Non-attendance may render the complication screening and management incomplete. Non-attendance is associated with significant opportunity costs in the health care systems, which acts as a barrier for healthcare access. Two studies in UK revealed that the non-attendance rate of diabetes care clinic was 15% (5) and 19.5% (6). One of the commonest reasons of non-attendance reported in literatures is simply forgetting the appointment (7)(8)(9)(10)(11). A local exploratory qualitative interview of the lower socio-economic class patients with diabetes found that they may have difficulties to read multiple appointments slips (12). Therefore, a patient reminder system before the appointment time could potentially help to improve the attendance rate.
Currently, patient reminder can be in the form of posted letters, telephone calls or electronic text messaging. All of them showed improvement in attendance rate in different clinical settings and populations (13)(14)(15)(16). Text messaging was found to be the most cost-effective method (17). Electronic text messaging could be delivered through Short Message Service (SMS) or through mobile applications in smart phones. Almost all mobile phones can receive SMS instantly with no additional charge, and it generally does not require additional mobile data subscription plan from the network service provider or installation of mobile apps. The possession of mobile phone by Hong Kong citizen was over 99% for aged 18-59 and 89% for aged 60 or above, while possession of smart phone was 96.1% for aged 18-59 but barely around 55% for aged 60 or above (18). Therefore, unlike the messages delivered through mobile applications, SMS could cover a broader population especially in elderly, making it a suitable strategy for delivery of reminder for patients with diabetes.
In a Cochrane review in 2013, there was evidence which showed text messaging reminders increase 4 the attendance of health care appointments, including two randomized controlled trials in primary care setting (19,20). However, the quality of studies was only low to moderate due to heterogeneity of methodologies and study populations (21). The service types, such as allied health or nursing services, were not separately analysed. The effect of such reminder in patients receiving multiple appointments was also unknown.
To provide holistic care to patients with DM, the government-funded primary care outpatient clinic in Hong Kong had introduced a risk assessment and management programme (RAMP) since 2009 (22). It comprises of various NPLM services with input from nurse case managers and allied health professionals including dietitians, physiotherapists and optometrists in primary care clinics. All diabetes patients are invited to attend RAMP once every 1-2 years. Thus, patients enrolled in RAMP programme normally receive multiple appointments with different dates and time. The nonattendance rates of different NPLM services before study implementation were 20% to 30% in 2015.
The main objective of the study is to determine the effect of using SMS reminder on the attendance rates of various NPLM services for patients with type II diabetes, including nurse diabetes complication screening (DMCS), nurse individual counselling (DMIC) and group education (DMGE), dietitian counselling (DC), physiotherapy session (PT), retinopathy screening by optometrist (OPT) and blood taking appointment (BT). Our hypothesis for the study was that for patients with type II diabetes with more than one scheduled services, there would be an increase in attendance rates for those who received SMS reminder compared with those who did not.

Methods
This study was conducted in a parallel randomized-controlled trial (RCT) design. The allocation ratio of The study was carried out in a community primary care clinic located in Kwai Chung, New Territories of Hong Kong from 2 Oct 2016 to 28 Feb 2017. It serves mainly the older patients, those from a lower socio-economic class and those with chronic illness. The majority of attendees are Chinese. There were 6 primary care physicians, 6 nurses, 1 physiotherapist, 1 dietitian and 1 optometrist regularly working in the study clinic. It provides around 55,000 doctors consultations and over 14,000 RAMP service attendances annually. During the study period, a total of 2,648 patients with diabetes attended for medical consultations in the study site. The cost of various NPLM DM services was the same at around USD 6.4 (HKD 50), except for blood taking which was free of charge.

Study populations
Subjects who fulfilled the following eligibility criteria were included: [1] aged 18 or older with the diagnosis of type II DM (with International Classification of Primary Care 2 nd edition coded T90), [2] possessed an SMS-capable phone, [3] could read Chinese or English, and [4] had at least 2 of the 7 NPLM appointments booked in the study site (DMCS, DMIC, DMGE, DC, PT, OPT, BT) scheduled 2 days or more following the recruitment date.
Subjects were excluded if: [1] their diabetes was managed by the internal medicine specialists or endocrinologists at the time of recruitment, or [2] they were mentally incapacitated.

Sample size calculation
The attendance rates for the RAMP services was estimated to be approximately 70%. A study done in a Malaysia clinic showed SMS reminder can improve the attendance rate by 11.5% (20). By deploying the SMS reminder service, the attendance rate was expected to be 81.5%. In order to have 80% power and 5% false positive error, a sample size of 217 appointments were needed for both intervention and control group. Assuming all subjects had 2 appointments, the number of subjects needed in each group was 109. After consideration of the attrition rate of 15%, we aimed to recruit 6 126 patients in each group, making 252 subjects in total.

Study pilot
A pilot trial was conducted involving 30 patients and the data was not included for analysis.
Optimization of recruitment logistics, standardization of interviewing techniques and refinement of SMS reminder wordings were done with all the co-investigators before the study began.

Subject recruitment
Subjects were approached by the clinic doctors or nurses during their clinic visits and informed consents were obtained. Trained nurses then filled in the questionnaire about their demographics and habit of using mobile phone. Subjects were then directed to registration for appointment booking and they received their printed appointment slips. Subjects in the intervention group received an additional SMS as reminder. Other clinical and laboratory parameters, such as blood pressure, body mass index, duration of DM, presence of microvascular or macrovascular complications, use of oral anti-diabetic agents, haemoglobin A1c (HbA1c) level and low-density lipoprotein (LDL) level were retrieved from the electronic health record database. Figure 1 showed the study flow chart.

Randomization process
Each subject was assigned an anonymous identification code after recruitment. Subjects and the referring personnel were blinded from the group allocation. Simple randomization using random sequence generation with variable block length was used for assignment to either intervention or control group independently by the biostatistician, thus keeping the number of subjects approximately the same.

Intervention -SMS
Subjects randomized into intervention groups received an SMS reminder 24-72 hours prior to the scheduled appointment. A list of subjects in intervention group was prepared by the biostatistician given to the health care assistants (HCAs) who were not involved in the recruitment process. SMS was sent though a secured web-based interface provided by CSL Mobile Limited, Hong Kong by a trained HCA manually. The scheduled reminders were counter-checked with our system by another HCA 1 day before sending to avoid input error. The delivery status was also checked to see if it was successfully 7 sent to the subject's mobile. As replying SMS may incur additional cost to them, communication through SMS was discouraged. A sample SMS reminder may read, translated in English, "Clinic Reminder: Your dietitian appointment is scheduled at 10:30am on 20/4/2017. Please call 3651 5411 if you cannot attend. Please do not reply the message. Thank you." As a safety measure, automatic forwarding of all incoming SMS to nurse-in-charge's mailbox was done to prevent missing important messages from our subjects.

Outcome
The outcome was assessed by investigators who were blinded to the group assignment. The primary outcome of this study is the attendance of the NPLM services. Attendance is defined as the patient registered for the service as recorded by the outpatient appointment system. Those who had rescheduled their appointment and attended were also counted as attended the service. For those appointments which were being cancelled before the scheduled time were excluded from the analysis.

Statistical analysis
Descriptive statistics were used to show the distribution of socio-demographic, occupational profile, lifestyle, clinical history, and to summarize the clinical data of the SMS intervention and control patients. Differences between groups were compared using chi-square tests for categorical variables and independent t-tests for continuous variables. The effect of SMS intervention on the attendance rates of different DM services appointments of subjects was assessed. The attendance rates of each DM services appointments between intervention and control groups were tested using Chi-square test. Additionally, sensitivity analysis was also performed comparing only the attendance rate of the first visit of each subject using Chi-square test or fisher exact test. Logistic regression was performed to estimate the odds ratios (OR) of the SMS intervention on attendance rate of different DM services appointments group relative to the control group. Stepwise logistic regression was also performed to adjust for confounding factors. 95% confidence intervals of attendance rates and odds ratios were calculated. All statistical analysis was performed using SPSS Version 24.0. All significance tests were two-tailed and findings with a p-value less than 0.05 were considered statistically significant.

Results
Throughout the 5-month recruitment period, 297 subjects were approached. 12 subjects did not fulfil the inclusion criteria (10 of them had only 1 appointment on booking and 2 did not have DM) and 3 were excluded as being followed up in medical specialists. 11 subjects refused to participate, making the overall response rate 96%. A total of 271 subjects were recruited and randomized. 135 were in the intervention group and 136 were in the control group. The Consolidated Standards of Reporting Trials CONSORT diagram is presented in Figure 2 (23).
The details of the socio-demographic variables, clinical characteristics and laboratory data for both groups were listed in Table 1. There were no statistically significant differences of all the variables between the intervention and control groups. The mean (± Standard deviation, SD) age of the subjects was 63.3±11.3 and male to female ratio was 0.99. All of them are Chinese and 62% of them had no full-time employment. 41.3% had their highest education attained at primary school or below.
More than three-quarters (77.5%) of them had been using mobile phone for more than 5 years, and 82.2% of them possessed a smart phone.
Majority of them (88.6%) were on at least one anti-diabetic medication. Their mean duration of DM was 6.9±5.3 years. Their mean HbA1c was 7.4±1.3% and LDL level was 2.4±0.7 mmol/L. More than half (51.3%) had at least one macrovascular or microvascular complications of diabetes.
Most of the subjects (86%) had either 2 or 3 scheduled DM-related services in our clinic. The mean time to appointment of various services were 48.6±28.7 days. The overall appointment rescheduling rate was 1.6%. The cancellation rate was 2.4%, accounting for some subjects only had one appointment for analysis ( Table 2).
The SMS delivery rate for the appointment in the intervention group was 100%. 733 appointments (intervention N=372, control N=361) were analysed. The attendance rates of various services in both groups are presented in Table 3. BT, DMCS and OPT were the three most frequently scheduled appointments in our study, constituting more than 80% of the total. The overall attendance rates were 88.0% (92.7% in the intervention group and 83.1% in the control group). The appointment with highest attendance rate was blood taking (overall 96.8% attended, 98.1% in intervention group and 9 95.5% in control group), while dietitian appointment was the lowest (overall 70% attended, 78.8% in intervention group and 59.3% in control group). DMCS, DMIC, OPT and overall appointment attendances reached statistical significance in univariate analysis by Chi-square test.
Sensitivity analysis was also performed by comparing only the attendance rate of the first visit of each subject (Table 4). 271 appointments (intervention N=135, control N=136) were analysed which showed there were significantly higher overall attendance rates in intervention group (96.3%) when compared with the control group (84.6%).
To determine the effect of the socio-demographic (age, sex, occupation, education level, smoking and drinking status, marital status and use of smart phone), clinical variables (presence of diabetes complications, use of oral anti-diabetes medications, use of insulin, HbA1c level, systolic and diastolic blood pressure, low-density lipoprotein level, body-mass index and duration of diabetes) and days to appointment on the appointment attendances, stepwise logistic regression was performed. HbA1c level, days to appointment and SMS intervention showed significant effect on appointment attendance ( Table 5).
The effects of SMS intervention on the appointment attendances by logistic regression is shown in Table 6. The overall appointment attendance was significantly higher in intervention group with adjusted odds ratio (aOR) 2.55 (95% confidence interval, CI 1.57-4.12, p <0.001). The absolute difference of the attendance rates is 9.6% (92.7% in intervention group and 83.1% in control group).
For analysis of individual services, DMCS (aOR 3.17, 95% CI 1.25-8.03, p=0.015) and optometrist appointments (aOR 2.58, 95% CI 1.00-6.69, p=0.05) had significantly higher attendance rates in intervention group. The effect of SMS on other services did not show statistically significant difference.

Adverse events during study
There were no reported adverse events during the study period, including wrong numbers sent, data misinterpretation, loss of privacy or message delivery failure. Two incoming messages were received, both were acknowledging the service with appreciation.

Discussion 10
The conduction of this study was based on current evidence that one of the major reasons for nonattendance was forgetfulness. This was the first RCT focused in patients with DM demonstrated positive effect of SMS on attendance rate of multiple NPLM services in a government-funded primary care clinic. The characteristics of the study population was similar to those patients with DM who were being followed up in primary care clinics in Hong Kong (24), and it could be generalized to other primary care clinics with similar patient profile. The observed high response rate and absence of adverse events indicated SMS is a useful technology for primary care patients with DM.

Effect of SMS on appointment attendance
The study showed that SMS reminder improved the overall attendance rate of NPLM services as compared with no reminder. Further analysis of each individual service showed the attendance rates of DMCS and OPT appointments had significant improvement by using SMS reminder. Although the attendance rates of other services did not reach statistical significance, they were consistently higher in the intervention group. The frequency of DMCS, OPT and BT were higher than others because these were the components for annual diabetes assessment according to the prevailing service protocol.
Other allied health services were provided to patients if they were newly diagnosed diabetes, or referred on discretion by the primary care physicians and patients' preference. The results from this study are in keeping with the phenomenon that forgetfulness is one of major reasons for nonattendance.
The baseline non-attendance rate of NPLM services, excluding BT, ranged from 16.9-40.7% in the control group, which was consistent with our baseline and other studies(5, 6). One interesting observation was the high attendance rate of blood taking appointment. Previous studies reported besides forgetfulness, negative attitude, perceived ineffectiveness, personal commitments and transportation costs may also contribute to the appointment non-attendance (7)(8)(9)(10)(11)(12). In our study, 95.5% of patients attended blood taking even though no reminder was given to them in the control group. We postulated this phenomenon was due to the perceived usefulness or effectiveness of the service to the patients with diabetes. Subjects may find blood taking is more important to them as it is an accurate reflection of the disease control and may influence the decision of medication adjustment by the physician. In addition, blood taking service is free of charge and this may also contribute to high attendance rate. On the contrary, dietitian counselling had the highest nonattendance rate, despite all other NPLM services charged the same. In Chinese culture, food is not solely served as a source of nutrition, but it also plays an important role in social function. Chinese usually eat together to maintain relationship and for information exchange. They like to eat in Chinese restaurants with relatives, neighbours and friends in the morning. Some patients may have anticipated difficulty in following the advice by dietitians, so they would choose not to attend the counselling service as a way of avoidance. In order to sustain the improved attendance, patient education on the importance of diabetic services is equally important in addition to the reminder services. As there was no subject interview conducted during the study, the exact reason remains to be elucidated.

Comparison with existing literature
The findings from our study were accordant with four similar RCTs comparing SMS effect on primary care clinics attendance rate (19,20,25,26). Their attendance rate in control group ranged from 48.1% to 80.5%, which was lower than that of our control group (83.1%). The effect of SMS ranged from increase of 3.7% to 11.5%, which is similar to overall increase of 9.6% attendance rate in our study. The subjects in previous studies did not focus on patients with specific disease. The analysis of different types of multidisciplinary service was not performed. Our study added value in showing there is a significant effect of SMS reminders for improving appointment attendance rates in patients with DM. They need long-term and multidisciplinary care, which patients may receive multiple appointments from different service providers. The study result also provided some insight in application of SMS reminders for patients with other chronic diseases who need multidisciplinary services.
Two other factors, HbA1c and days to appointment, were found to have significant negative effect on appointment attendance. A recently published systematic review showed that the poor disease control with high HbA1c level was associated with follow-up non-attendance, which was consistent with our findings (27). It was also observed in our study that longer days to appointment was associated with lower attendance rate, and similar finding was also reported in a cross-sectional study in a ophthalmology clinic (28).

Limitation of study
As limited by manpower and available facility, the RCT was carried out in one clinic. This study included subjects who had DM and followed up in public sector, the effectiveness of SMS in other settings (e.g. those who had follow up in private sector) may require further evaluation. As the subjects were recruited based on the number of appointments but not the service types, some services may not have enough power to perform subgroup analysis. Also, we did not require the subject to acknowledge our SMS message after reading. Some of the subjects may have received but not read the reminder. It may underestimate the effect of SMS on the attendance rate. Lastly, health outcomes such as HbA1c, blood pressures and lipid profile were not assessed after the intervention.

Implication for clinical practice and future research
This study found that SMS is an effective strategy to improve the attendance rates of various NPLM DM services, particularly for DMCS and OPT appointment. With further development of this reminder system, for example with automatic transmission of reminder by connecting to the appointment system, the labour cost and the data input could be further reduced. The cost for each SMS is HKD 0.6 (equivalent to USD 0.08). It can reduce the wastage of scarce medical resources by patients' nonattendance. Since the process involves handling of sensitive personal data, health care providers need to consider possible concerns for patients' confidentiality and privacy issues when launching it as a long-term routine service model. Future research could be done to evaluate the cost-effectiveness of using SMS reminder and whether there is any difference in long term health outcome. Qualitative interview with patients who received SMS but still did not attend the services would have given insight into other reasons of nonattendance. The effectiveness of the SMS with variation on the timing, frequency and the content of the reminder message could also be studied. Under the current scheme, the SMS reminder is unidirectional (i.e. from clinic to subject). It is useful to explore the feasibility of a bidirectional communication using the same platform. 13 Conclusion SMS reminder is an effective way to improve overall attendance rates of non-physician-led DM-related services, particularly for diabetes complication screening clinic by nurse and retinopathy screening by optometrist. To reduce resource wastage, policy-maker could consider implementation in a larger scale in outpatient setting.