| |
Epi Wit & Wisdom Resources
Computer
Assisted Telephone Interviewing (CATI) Software (1 of 2)
By Tamar Lasky,
PhD
The School of Medicine at the
University of Maryland, Baltimore has established a Computer Assisted
Telephone Interviewing (CATI) facility as part of the Center for
Health Policy and Health Services Research. Our experience with CATI
suggests new applications and research opportunities to share with
colleagues.
Our CATI facility uses Info Zero
Un software for telephone surveys that we conduct for investigators
throughout the School of Medicine, for other professional schools
within the University of Maryland, and for outside contractors. The
software was purchased after comparison with other software packages
(the runner up was Sawtooth software) and selected because it offered
more versatility it could run networked or stand alone, and the
hardware requirements are minimal. We have used the software since
January 1997 and have grown more and more enthusiastic about its value
to epidemiologists and its potential for applications beyond data
collection. It expands our questionnaire design capabilities, greatly
reducing the administrative and quality control burden related to
telephone surveys, systematizes telephone calls, and has the potential
to be used to improve patient compliance with treatment or in clinical
trials, and to affect the provider/ patient relationship. Some of our
comments below apply only to the Info Zero Un software (the basis of
our experience), but much of what we have observed can be generalized
to other CATI technology. The Info Zero Un software (a French name,
because it was created in Montreal) includes several capabilities:
programming the questionnaire, programming interviewer parameters,
supervision and monitoring interviews, and statistical analysis.
Programming the
Questionnaire
Many of us have used packages
such as Access or EpiInfo to design and administer questionnaires.
These packages contain some of the capabilities of CATI software such
as creating skip patterns, limiting ranges of response values, and
handling open-ended questions. However, Info Zero Un (and some other
CATI packages tailored for telephone interviewing) go further,
permitting: 1) randomization of question blocks so that responses are
not biased by the order in which they appear in the interview; 2)
randomization of response options so that responses are not biased by
the order in which they are offered to the respondent; 3) look up
tables embedded in the questionnaire for coding (i.e. ICD-9 codes,
occupational codes, census tracts, etc.); 3) calculations that can
lead to logical skips (i.e. x number of glasses equals yy mg of a
nutrient and respondents over or under specified values get skipped to
appropriate questions); 4) elimination of response choices at a given
question, based on responses to previous questions. These and other
capabilities open up new possibilities in questionnaire design and
even suggest new types of research questions. Question patterns that
were confusing or too complex with other software become easily
managed with specialized software.
Programming Interviewer
Parameters
Most CATI software provide
capabilities for programming and scheduling telephone calls, call
backs, appointments, and other rules affecting the methodology of the
telephone interview. In the software we used, hours of telephone calls
can be programmed for each project and interviewer taking into account
the hours of operation of the unit, as well as the desired hours for
conducting interviews, and the time zones in which the respondents
live. The number of attempts is programmed, as well as rules
governing each attempt. For example, after the first busy signal, one
might program the system to call after 10 minutes, and if busy then to
call in 30 minutes, and if busy then to call the next day. One can
program each call to count as an “attempt”, or one could consider the
two calls 10 minutes apart to be one “attempt.” The programming
insures that the same number of attempts are made to reach all
potential respondents and that the attempts are made according to the
same set of rules.
Supervision and
Monitoring of Interviews
Those of us who have had
administrative responsibility for telephone surveys will greatly
appreciate the record keeping handled by software designed for CATI.
Such software collects information on each attempt, reasons for
refusals or unsuccessful attempts, length of interviews, problem
questions and interrupted interviews. Data and reports may be
generated for each day or time period, project and interviewer. The
system permits a supervisor to monitor interviews on their own
computer in a separate office or from home, and to e-mail comments to
the interviewers during their interviews.
Statistical Analysis and
Report Generation
Reports about administrative
aspects such as number of interviews and length of interviews can be
immediately accessible, as are frequencies of responses and printouts
of open-ended responses. Data collected by Info Zero Un are easily
generated as SAS or SPSS data sets for additional statistical
analysis.
Implementation
The system can be used by a sole
investigator or provider, but maximal efficiency may be achieved when
established as a central facility. Our facility is structured in three
tiers: 1) a PhD epidemiologist providing scientific guidance in survey
design, questionnaire design, sampling strategies, data quality and
validity; 2) a research coordinator/project manager to program the
survey, supervise interviewers, and produce reports; and 3) telephone
interviewers whose work is monitored for quality and productivity. The
software and training to operate one station can cost several thousand
dollars, a cost that may be balanced in a medium size study by savings
in data entry and administrative record keeping. The cost of running
more than one station is increased by the need to network stations
(not a necessity, but a big advantage) and for staffing an
infrastructure to train and supervise interviewers, prepare reports,
and coordinate multiple projects. In addition to software, we
purchased chairs and computer furniture recommended by a physician
with expertise in occupational medicine and by a physical therapist as
designed to minimize the development of job-related pains and
injuries. Total expenditures for hardware, software, furniture and
headsets for a five station facility could be about $25,000, varying
with the equipment selected. This cost may be reasonable for a large,
multi-center study; again data entry costs are eliminated, as are many
administrative hassles and costs. I believe that data quality is
improved, primarily because interviewers are not distracted by the
need to record administrative details, calling methodology is
monitored and systematized, and interviewer productivity and quality
data are easily generated.
Although the system requires a
fair amount of technical knowledge, a research coordinator with
minimal computer skills can acquire the expertise needed to run the
system upon completion of the company’s three day training course. My
staff received software training in January 1997, and began survey
work in April 1997. Between April 17 and August 8, 1997, we completed
three telephone surveys involving 25,334 attempted telephone calls in
an effort to reach 5,848 persons, and completion of 675 interviews.
Fifteen hundred hours of interviewer time (seven individuals working
up to 32 hours/week) included time spent on training for each project.
Patient compliance and
enrollment Computer Assisted Telephone Interviewing (CATI) systems are
generally used to collect data in a one time unidirectional flow, but
their increasing flexibility and versatility suggest applications in
settings where information flows in two directions: 1) the telephone
interview can be used in the traditional capacity to gather
information about large groups of people and 2) the telephone
interview can be part of an information flow between the researcher or
health care provider and the patient to gather information, respond to
that information in a timely manner, and to bring a message to the
patient. The message may be a generalized message of concern about the
patient, expressing receptivity to the patient’s needs, and concern
for the patient’s well-being, and the message may be specific,
reminding the patient about clinic appointments or treatment issues.
In this situation, CATI can be used to improve the relationship with
the patient, encourage medical compliance, and encourage participation
in a research study.
|
|
