JUSTIFICATION
The Outpatient Department and the Medical Informatics Department of the Hospital Italiano de Buenos Aires (HIBA) designed a joint strategy based on e-mail reminders for scheduled outpatient appointments. Patients who have an appointment (either face-to-face or teleconsultation) scheduled with a doctor at the Hospital receive an e-mail at the time of their request as well as 14 days and 48 hours before the appointment, reminding them of the appointment time and the place where they have to go. In addition, a button has been added to the e-mail so that the patient can click on it and cancel it directly in a simple way.
The possibility of canceling the appointment in the same reminder is also intended to have an impact on those who, without having forgotten their appointment, have to or want to miss it, can cancel it and thus leave the appointment free for another patient who needs it.
It seems that any type of reminder would be associated with a decrease in the no-show rate. However, it is questionable whether published figures are transferable to our setting. The variability of contexts and populations can affect the magnitude of the effect of an intervention.
After reviewing the literature, no studies were found analyzing appointment reminders and no-show rates in our country, so the investigators propose to conduct a study evaluating this impact in the Urology Department of HIBA. Although a clinical trial could be considered and the sending of reminders could be randomized, by the time the investigators decided to carry out this research, the intervention had already been implemented in all departments.
RESEARCH QUESTION
Is the e-mail reminder of a scheduled outpatient appointment to the Urology Department of HIBA effective in reducing no-show rates in patients over 18 years old?
SELECTION AND SAMPLE SIZE
During the year 2022, the Urology Department of Hospital Italiano de Buenos Aires (HIBA) performed approximately 29000 appointments in the Central Hospital and 11500 appointments in the Italian Hospital of San Justo, giving a total of 40500 appointments including spontaneous overtime and Spontaneous Demand appointments. The estimated weekly average is 779 appointments. Currently, in the Urology Service of HIBA, there is approximately a 21% no-show rate in scheduled outpatient appointments.
The investigators believe that achieving a 25% relative reduction in the proportion of no-show rates through the reminder system would be clinically relevant, i.e. from a 21% baseline no-show rate to 16%.
Using both STATA and the sample calculation tool provided free of charge by the University of San Francisco, the investigators estimated that the total N needed to estimate the calculated difference in no-show rates between groups is 1892 appointments (946 appointments per group). The parameters used were as follows: alpha (two-tailed)= 0.05, Beta= 0.20; q1 (proportion of appointments exposed)= 0.50; P0 (baseline risk of no-show rate in the unexposed group)= 0.21 and P1 (estimated risk in the exposed group)= 0.16.
The number of appointments requested before and after the implementation of the reminders is enough to test the study hypothesis.
DATA COLLECTION INSTRUMENT
For our study, the investigators will use secondary databases requested from the Department of Health Informatics after approval of the protocol by the bioethics committee of our institution. Because appointment management is centralized in a single data repository, all data corresponding to patients and appointment requests can be obtained from the entire HIBA electronic health system.
PLAN FOR DATA ANALYSIS AND REPORTING OF RESULTS
An individual approach will be used to evaluate the effect of the intervention on no-show rates. The unit of analysis is the assigned appointment. Due to the nature of the exposure and the implementation of the reminder system, all appointments before the implementation date correspond to unexposed appointments, and all appointments assigned after the implementation date correspond to exposed appointments.
Quantitative variables will be described with their absolute and relative frequency in percentage. Quantitative variables will be presented as mean and standard deviation or median and interquartile range according to the observed distribution.
For the calculation of the rates of no-shows and canceled appointments, the count of no-shows or canceled appointments, respectively, will be used as the numerator. For the no-show rate, the denominator used is the number of appointments not canceled, as is common for presenting this indicator in the literature. In the rate of canceled appointments, the total number of requested appointments will be used as the denominator. In both cases, only appointments that meet the inclusion criteria and none of the exclusion criteria during the same period will be used. The 95% CI (Confidence Interval) estimated with the normal approximation will be presented for each of the rates because the proportions are sufficiently large.
To test the null hypothesis of equality in the proportion of missed appointments between requested non-canceled appointments exposed to the reminder and requested non-canceled appointments not exposed to the reminder, a multivariate logistic regression model will be used. The investigators consider exposure to the reminder as the main exposure and no-show to the appointment as the outcome variable. Each patient will be considered as a conglomerate or cluster using robust estimators of the standard errors since the appointments of the same patient are more similar to each other than to the appointments of other patients. Time in days will be used as the adjustment covariate. The investigators will adjust for potential confounders considered relevant by the research team and the literature, which include sex, age, and medical coverage of the patient, time since the appointment request in days, time, day of the week and month of the appointment, place of care (Central Hospital, San Justo or Peripheral Center). The crude and adjusted ORs for missed appointments will be estimated with their 95%CI.
To test the null hypothesis of equality in the proportion of canceled appointments between requested appointments exposed to the reminder and canceled appointments among requested appointments not exposed to the reminder, the same methodology described for the previous objective will be used.
The effect of the intervention will be evaluated in pre-specified subgroups according to sex, age group, medical coverage, appointment time, day of the week, and month of the appointments, as well as according to place of care (Central Hospital, San Justo or Peripheral Center).
Probabilities of less than 0.05 will be considered statistically significant. Statistical analysis will be performed using STATA version 17 software (Texas USA).
HANDLING OF LOST DATA
The databases to be used in this analysis will contain a very low percentage of missing data. Because any missing data could be considered as missing completely at random, a complete case analysis will be used for the individual analysis.
BIASES AND LIMITATIONS OF THE STUDY
The limitations of our study, given the chosen methodology, will be mainly due to the lack of randomization of the exposure. This leads to confounding factors, which, although some will be considered a priori, cannot be considered in their totality since not all of them are known. On the other hand, the sample groups may not be balanced in terms of confounders, as may happen, for example, with time-dependent variables. However, since the intervention was implemented from a randomly defined point in time, the confounders are most likely to be balanced and, in addition, can eventually be adjusted through statistical methods.
While many retrospective studies are conducted based on incomplete or inaccurate information, many of the administrative variables in our study will be collected completely and automatically, with a standardized process that will remain stable throughout the study period. Unlike the administrative variables in the study, other variables, such as date of birth, may contain errors or incomplete data because although they are computerized, the data are originally uploaded by hospital administrators who are prone to error.
An additional limitation is all the unmeasured variables such as socioeconomic status and/or eHealth literacy of the individuals. Not all people open their e-mail every day and by not being able to measure who read the reminder the investigators may falsely believe that the intervention did not meet its objective when in fact what happens is that the patient did not open the e-mail and so was not exposed to the reminder. The investigatorsconsider it interesting in the future to compare our results with a reminder sent directly to the patient's cell phone via text messages or WhatsApp.
On the other hand, although the implementation of the intervention does not incur an additional cost to the institution for the aforementioned reasons, the investigators will not evaluate whether the intervention obtains any economic return.
Another limitation of our study is that the investigators will not evaluate the acceptability of the reminder by the users. If given a choice, it is possible that patients would choose other types of reminders according to personal preferences.
Nor will the investigators be able to estimate the impact of the intervention regarding teleconsultations, since these will be excluded from the analysis.
As far as external validity is concerned, the results obtained may not be extrapolated to other healthcare centers that do not have the structure of our hospital. To send e-mail reminders to all patients, it is necessary to have an electronic system that combines and processes information from the electronic medical record and the scheduling system and to coordinate it with a mass mailing system to be able to do it automatically. Otherwise, it is necessary to have human resources to do all this, which may not be cost-effective.
Last but not least, it is worth mentioning that since the intervention studied has two components (the reminder itself that prevents forgetfulness and the facilitator of appointment cancellation), it is not possible to estimate the effect of each component of the intervention separately.
EXPECTED IMPACT
The intervention is expected to decrease the no-show rate by 25% (or, in other words, to bring the no-show rate down to 16%) and the average waiting time for an appointment by at least one day. In addition, it is expected to increase the number of cancellations by 20%.
RESOURCES AND ESTIMATED BUDGET
The needed resources are available to complete this work. In no case will this study represent any additional cost for the patients or their medical coverage. The HIBA Medical Informatics Department already has a mass mailing system, so sending reminders does not require any financial investment or additional human resources.
The team in charge of the research and in charge of searching, processing, and analyzing the information has enough protected time during the working day to be able to complete the study within the stipulated deadlines.
ETHICAL CONSIDERATIONS
This study protocol has already been approved by our Hospital's Bioethics Committee: IRB (Institutional Review Board) Nº 6878
