Using Telephone Prompts to Improve Initial Attendance at a Community Mental Health Center

Subjects were all persons newly referred to Hillview, a publicly funded community mental health center in Wellington, New Zealand, between November 1991 and January 1995. Those with any previous contact with the center were excluded to reduce confounding effects of past experiences with the service.

At referral all patients were assigned a registration number. The experimental group consisted of those assigned even registration numbers, and the control group were those assigned odd numbers.

Appointment times were scheduled from one to eight weeks ahead. An appointment card was mailed to all patients within a week of the referral telling them when their appointment was. For the control group, no other reminder was given. The day before the appointment, clerical staff telephoned patients in the experimental group during normal working hours to remind them of their appointment.

For the telephone reminders, staff attempted to speak with the patient in person. For reasons of confidentiality, they did not leave messages. They recorded the number of attempts to make contact, whether contact was achieved, and whether the patient attended, failed to attend, or cancelled the appointment.

For all subjects, age, gender, referral source, identified problem, waiting time to appointment, and whether the patient had provided the mental health center with a contact telephone number were determined by retrospective case file review.

Data were analyzed using SAS. Logistic regression was used for the analysis. First, characteristics of the experimental and control groups were compared; they included age, gender, referral source, type of problem, waiting time, and whether the patient or referrer had provided a phone number to the mental health center. These factors have been identified as possibly influencing attendance. We expected that because of the random allocation of patients between groups, no differences would be found on these variables.

Second, the effect of the phone call on patient attendance was modeled, and odds ratios were calculated to measure the size of the effect. In addition, the effect on attendance of the characteristics listed above was investigated. The final analysis examined how patient characteristics may have influenced the effect of the phone call on attendance.

A total of 1,087 new patients were referred during the study period. Thirty-four were redirected to other services, and data were incomplete for 77 patients because of administrative oversight unrelated to any patient factors. Excluding these patients resulted in a total of 976 subjects—480 in the experimental group and 496 in the control group. The difference in group size is due to the fact that the excluded patients were not evenly distributed between the two groups.

The mean±SD age of the patients in the experimental group was 33.65± 11.02 years, with a range of 16 to 82 years. A total of 339 patients in this group (71 percent) were female. For the control group the mean±SD age was 35.29±11.91 years, with a range of 16 to 75 years. A total of 333 patients in this group (67 percent) were female. The two groups did not differ significantly in age or gender. Data on race were not routinely recorded by staff at this stage of the treatment process, and this information was not available for analysis.

Because of unanticipated staff shortages, an attempt was made to contact by phone only 262 patients in the experimental group; thus for 45 percent of the experimental group no contact was attempted. Of the 262 patients for whom a phone contact was attempted, no details about the contacts were documented for 39 patients, and it was not possible to speak in person to 33 patients.

The only significant difference in baseline variables between the experimental and control groups was that a larger proportion of patients in the experimental group did not provide the mental health center with a telephone number, 92 patients, or 19 percent, of the experimental group compared with 67, or 14 percent, of the control group (χ²=5.419, df=1, p=.02). No significant differences were found in referral source or identified problems.

A total of 182 patients contacted by phone (96 percent) attended their appointment, compared with 375 patients in the control group (76 percent), and 50 patients for whom telephone contacts were attempted but were unsuccessful (70 percent). The latter two groups did not differ significantly.

Successfully contacting patients by telephone increased the likelihood of their attending the initial appointment by a factor of 4 (OR=4.2, p<.001). The rate of attendance for patients in the experimental group for whom telephone contact was not attempted due to staff shortages was significantly lower than the rate in the control group (53 percent versus 76 percent; OR=.3, p<.001).

In both the experimental and the control groups, patients between the ages of 15 and 25 were significantly less likely to attend than patients in other age groups (OR=2.2, p=.01). Those with a waiting time of more than 21 days until their appointment were significantly less likely to attend than those with a shorter waiting time (OR=2.8, p=.002). Patients who provided the mental health center with a telephone contact number were more likely to attend than those who did not (OR=1.6, p=.03). After these variables were controlled for, patients in the group that received telephone reminders were still four times more likely to attend than patients in the control group (OR=4.4, p<.001).

Gender, referral source, and identified problem did not have significant effects on attendance.

The results confirm previous findings that attendance rates can be dramatically improved by telephone reminders within 24 hours before an appointment. Our results are strikingly similar to those of Hochstadt and Trybula (5), who found an attendance rate of 91 percent for patients who were contacted by telephone within 24 hours of their appointment, compared with 68 percent for those who were reminded three days beforehand by telephone or letter and 45 percent for those who were not reminded.

However, the study reported here also demonstrated that telephone prompting is not a straightforward matter. Due to staff shortages, more than 200 patients in the experimental group were not contacted by telephone. This uncontacted group had an unexpectedly low rate of attendance (53 percent) compared with the rate for the control group (76 percent). This group also had the highest rate of cancellations (15 percent versus 8 percent for the experimental group and 10 percent for the control group).

A similar finding in Hochstadt and Trybula's study (5) was not explained. Those authors reported a 33 percent nonattendance rate over a 27-month period before their study and a 55 percent nonattendance rate during the study for the group who did not receive a reminder.

In our study, failure to attend was not distributed evenly over the year but had seasonal peaks, at Christmas, during school holidays, and in midwinter. These peaks coincided with times when staff failed to attempt telephone reminders, because the administrator was on holiday or overwhelmed with seasonal pressures or by illness. This pattern may explain why the nonattendance rate was higher for patients in the experimental group for whom contact was not attempted; these appointments were concentrated in the seasonal periods of maximal nonattendance. Also, a smaller proportion of the experimental group provided telephone contact numbers to the mental health center. However, this factor was controlled for in the final analysis.

Data collection in this study was limited because data were gathered by staff who were already fully committed to clinical duties. Their other time commitments meant that some experimental subjects were not contacted by phone and that some types of data were not collected, such as information about ethnicity and socioeconomic status. Despite these limitations, the study showed that contacting patients by telephone within 24 hours before their initial appointment dramatically improved attendance rates in a public health setting in New Zealand. Further, the findings highlight the practical difficulties of contacting patients, especially when staff time is scarce. In health systems that increasingly focus on efficiency, the findings illustrate the false economy of saving costs by reducing the number of clerical staff.

The authors are grateful to Robyn Green, B.Sc., Dip.O.R.S., for statistical assistance; to Marg Grootelaar for administrative work; and to the Medical Council of New Zealand for financial support of Ms. Brown's work.

When this study was done, Dr. MacDonald was affiliated with Hillview Community Health Team in Wellington, New Zealand. She is now senior lecturer in the department of psychological medicine at Wellington School of Medicine at Otago University, P.O. Box 7343, Wellington, New Zealand (e-mail, macdonaldj@wnmeds. ac.nz). Ms. Brown is with the department of veterinary science at Massey University in Palmerston North, New Zealand. Dr. Ellis is professor and head of the department of psychological medicine at Wellington School of Medicine at Otago University.