The American Journal of Managed Care July 2008
Automated Patient Assessments After Outpatient Surgery Using an Interactive Voice Response System
Objective: To test the feasibility and utility of an interactive voice response system (IVRS) for monitoring patients after outpatient surgery.
Methods: We studied consecutive patients undergoing gynecologic day surgery. The IVRS called patients on the first postoperative day and asked them if they were experiencing new problems. Feasibility was assessed in terms of call responses and acceptance by patients. Utility was measured in terms of the ability of the IVRS to identify adverse events (AEs), defined as procedure-related symptoms requiring a physician or hospital visit. We contacted patients 30 days later to elicit their perceptions of the IVRS and determine AE status.
Results: Follow-up was complete for 249 of 270 enrolled patients (92%). The IVRS successfully contacted 130 patients (52%). Of the 22 patients (17%) who required a follow-up phone call, 9 had a new problem related to surgery, 7 had new or worsening symptoms, 6 wanted to speak with a nurse, and 1 had a medication-related problem. Patients remembering the automated call (n = 96) reported the system easy to use (82%) and comprehend (86%). Most patients (68%) preferred the IVRS to a personal follow-up call (probability greater than 50%, P <.001). AEs occurred in 40 patients (16%; 95% confidence interval = 12%, 21%). The IVRS did not identify any AEs because 90% of these occurred after the automated call.
Conclusion: An IVRS-based method of monitoring outpatient surgery patients is feasible. To improve utility, calls must occur later than first postoperative day.
(Am J Manag Care. 2008;14(7):429-436)
To monitor for adverse events in patients undergoing outpatient surgery, many institutions have a policy for nurses to call patients on the first postoperative day, a process that is expensive and difficult to audit.
An interactive voice response system (IVRS) can be used for postoperative calls.
Use of the IVRS is more efficient and does not compromise the quality of the assessments.
However, to improve utility, calls must occur later than the first postoperative day.Patients who undergo outpatient surgery do not stay in a hospital facility the evening after their procedure. With advances in minimally invasive surgery and anesthetic techniques, the volume of outpatient surgery is increasing dramatically and accounts for the majority of surgeries.1,2 Although outpatient surgeries are generally safe, risk is increased in certain circumstances.3-5 Outpatient surgery complications can be especially damaging because they occur at the patients’ home outside of a monitored setting.
Fortunately, most complications occurring after outpatient surgery are self-limited. The most common complications include symptoms such as pain, nausea and vomiting, bleeding, and headache.6 Although these symptoms often are of minor clinical importance,4,6,7 it is important to monitor them for 3 reasons. First, symptoms can herald severe underlying problems such as wound infections or postoperative myocardial infarctions. Second, these symptoms can progress to severe outcomes such as shock, respiratory failure, or death, especially if there is a delay in treatment.8,9 Finally, symptoms can cause anxiety and impede a rapid return to normal function.
One common method of assessing complications after outpatient surgery is a nurse callback program. In such programs, a nurse calls patients on the day after surgery and will subsequently coordinate treatment if a patient is experiencing problems.10-12 A limitation of such programs is their expense. To reduce costs, one could automate aspects of it using an interactive voice response system (IVRS),13 which is a technology allowing patients to interact with databases using a telephone.14-21 The IVRS could make an initial screening call to all patients to determine whether they were having a problem. Then, the IVRS could notify a nurse to follow up with the few patients who were screened “positive.” This practice would obviate the need for a nurse to call every patient, significantly reducing the monitoring program’s expense.
Two issues need to be considered by decision makers when contemplating whether to implement an IVRS-based patient monitoring system. First, will patients accept the technology? If patients cannot use the IVRS, then the technology is not capable of improving monitoring. Second, will the IVRS-based monitoring result in improved identification and management of complications? If the system does not alert providers to patient problems in a timely manner, then the system is not useful. We designed this study to assess these factors. We were primarily interested in understanding issues related to feasibility to determine whether further study of IVRS to manage patients was merited. We also were interested in understanding the potential utility of IVRS in terms of managing complications, particularly its ability to capture the timing and types of outcomes patients experienced. Because these data were not previously described, this study was necessary for the design of future IVRS-based interventions.
Design, Setting, and Patients
We used a prospective cohort design to study women undergoing outpatient gynecologic surgery at the Royal Alexandra Hospital in Edmonton, Alberta, Canada. The Royal Alexandra Hospital is a teaching hospital within the Capital Health Region of Alberta. It has an active outpatient surgery program, conducting more than 14,000 procedures annually. Consecutive patients undergoing surgery between April 19, 2006, and June 9, 2006, were eligible. We excluded patients if they did not have a telephone, did not speak English, had cognitive impairment, or did not provide informed consent. We enrolled patients on the day of their surgery as they waited for their procedure. The study was approved by the joint University of Alberta/Capital Health Ethics Review Board.
The study intervention is illustrated in Figure 1. After patient consent, we collected demographic and medical data. When patients were discharged from the outpatient surgery unit, the study nurse entered the patient’s ID, phone number, and discharge date into the IVRS; this information was physically stored on a server in the hospital’s data center. The nurse performed data entry using a Web-accessible personal computer in the outpatient surgery unit. The IVRS called patients on the day after their discharge from the day surgery unit. If there was no response to the call, the IVRS reattempted the call every 2 hours until a response occurred. If no response occurred before 9:00 pm that day, the system stopped calling the patient.
When a call was answered, the IVRS asked a series of questions that required a YES/NO answer. The IVRS used speech recognition software to update the database and prompt further actions. The system first asked: “Are you the patient discharged from the Royal Alexandra Hospital yesterday?” If the person responded NO, then the IVRS thanked the individual and disconnected. If the person responded YES, IVRS asked 4 questions and stored the responses:
• “Since your surgery, have you had any new or worsening symptoms?”
• “Since your surgery, have you had any problems related to your surgery?”
• “Since your surgery, have you had any problems related to your medications?”
• “Would you like to speak to a nurse?”
The IVRS transferred the patient to the provincial teletriage program (called HealthLink) after their first YES response to 1 of these 4 questions. HealthLink is a call center with nurses available to respond to patient queries at all times. Calls are answered by nurses, who are trained to triage patients according to the patient’s problem and prespecified call algorithms. Call center nurses have specific algorithms to deal with postoperative problems.
The IVRS we used was CallAssure. CallAssure can be run on any modern personal computer equipped with a telephony card and at least 2 analog lines. In addition, CallAssure requires access to an e-mail server (for notification messages) and, optionally, a printer (for reports). CallAssure uses Nuance 8.5 as its speech recognition software. Nuance reports an overall accuracy of 97%, but this rate is likely much higher for responses to questions requiring YES/NO answers.22 We did not use routines for dealing with grammar mismatch errors to keep the system as simple as possible. Also, in previous testing, we discovered that patients found it difficult to revert to touch-tone inputs as many people use handheld devices. In our experience, these devices made it very difficult to use touch-tone input.
Follow-up and Study Outcomes
Thirty days after hospital discharge, we manually contacted all patients by telephone to administer a telephone survey. This survey was completed independent of any knowledge pertaining to the IVRS call. Patients were asked their recollection of the automated call, their opinion about some aspects of the IVRS (eg, perceived usefulness), whether they had any new or worsening health problems after their surgery, and whether they required any visits to providers since their surgery.
We supplemented the telephone survey with information from the health region’s administrative data and the Health- Link call logs. We searched the respective data systems for visits to hospitals, results for any laboratory tests, or calls to HealthLink by the study patients within the 30 days after the surgery date. When such encounters existed, we obtained relevant information from the system, including the date, time, and reason for the hospital visit; the specimen type and date, as well as the results for all laboratory tests; and the call dialog, date, and time of all calls to HealthLink.
This study’s main goal was to assess the IVRS feasibility, which was done by measuring the IVRS’s call statistics and the survey results. A secondary objective was to determine the potential utility of using an IVRS for monitoring patients for adverse events. The relevant outcome for this aspect of the study was an adverse event, which we defined as any procedure-related symptom or poor health outcome requiring a visit to a physician or hospital.
Statistics and Data Management
We used SAS version 9.1 (SAS Institute, Inc, Cary, NC) for all data analysis. We report descriptive statistics using median and interquartile range (IQR) for continuous variables and frequency distributions for categorical variables. For outcomes, we present the proportion of patients with the outcome of interest with 95% confidence intervals (CIs), which we calculated using the Wilson score method.23
We used a multivariable logistic regression model to measure the association between IVRS call response and the likelihood of experiencing an adverse event while controlling for patient and surgery factors. We first measured the association of experiencing an adverse event with patient and surgery factors at the univariate level by using the χ2 test or the Wilcoxon rank sum test, respectively. We entered variables into our multivariable model if they were statistically associated with adverse event status at the P <.20 level.
Role of the Sponsors
The sponsors did not have any role in the design, conduct, analysis, or write-up of the study.
Figure 2 summarizes the study flow and the main results for all patients. We enrolled 270 of 312 eligible patients during the study period. Thirty-four patients were not enrolled because they were missed by the research nurse, as she could not approach all patients. Eight patients refused participation in the study because they did not wish to take part in a research project. Of the 270 enrolled patients, 4 were not entered into the IVRS because of administrative errors and another 17 were lost to follow-up (6%). Our results are based on the 249 (92%) patients for whom we have 30-day follow-up data.
Table 1 describes the characteristics of patients in our cohort. The median age was 38 years (IQR = 31-49 years).
Most women had no comorbidities apart from the indication for surgery, virtually none had functional disabilities affecting their ability to perform activities of daily living, and few had surgeries in the preceding 12 months. The vast majority of patients were discharged on the day of surgery. The most common surgical procedures involved an abdominal approach (eg, laparoscopic surgery for endometriosis) or deep perineal approach (eg, dilation and curettage, cone biopsy).