What are some of advantages and disadvantages of computer aided telephone interviewing?

Telephone Surveys

Telephone surveys are structured interviews conducted by an interviewer who records the answers to mostly closed-ended questions. Improvements in electronic technology, particularly computer-assisted telephone interview (CATI) systems, have enabled interviewers to enter responses immediately into databases that commence aggregation and analysis of the entered data. In addition to this advantage, telephone interviews are relatively inexpensive to administer (unless calling long distance) and of short duration, compared to the time taken by respondents to complete written questionnaires. Disadvantages of telephone interviews include the limited range of times available to collect data, the potential expense of long-distances, and the limited ability to use open-ended questions.

Insufficient Data of Nonofficial Sources Arising from a Diversity of Statistical Classifications and Altering Collection Methods

Numerous nonofficial sources provide aggregated data on old and new communications and media markets, as well as microdata on communications and media enterprises, on individual media use and – on a minor scale – media content. Many of these data providers only offer data on a national level using their own statistical classifications. This leads to severe constraints for secondary analyses in social research.

One methodical problem arises from the tendency of industry associations and commercial data providers to change their classifications and data collection methods very frequently. This means that one has to be careful when using data from nonofficial statistical sources for time series analyses, even on the national level. A break in a statistical trend may be a statistical artifact due to altered statistical categories or data collection methods, and not the result of changing behavior of providers or users of communications and Internet services or media content. In 2000, for example, the interview technique of Germany's largest annual audience survey (media analyze) changed from personal face-to-face interviews to computer-assisted telephone interviews (CATI). A sharp rise in the level of radio listening time per adult of nearly 30 min per day was the result.

The different national statistical categories and collection methods also have a negative effect on the usability of the data for cross-country studies. Supranational nonofficial sources that claim to provide harmonized international data, by altering the original national data, have to be checked thoroughly. One example is data published on the time spent on different media. It is collected by either diary methods, personal interviews, or telephone interviews. There is clear evidence that the chosen method has an impact on the results. Another good example is the above-mentioned annual WARC report on international advertising markets. Some national advertising associations collect the so-called ‘net’ advertising turnover figures (i.e., without deductions and commission charges) directly from media companies. In other countries, commercial companies measure advertising volume (number of advertising spots or ad pages) and estimate ‘gross’ advertising expenditures by multiplying the volume figures with the official media tariffs, announced by media companies. However, in reality most of the advertising time and space is sold much cheaper. To harmonize data from different countries, WARC estimates the gross–net difference, but is not very successful in doing so. (In Germany, for example, both methods to collect data on advertising expenditures are in use. In 2012, the gap between published ‘net’ TV adverting figures and published ‘gross’ TV adverting spending was 70% while the estimate of WARC was 20%.) To avoid the production of statistical artifacts, all cross-country studies using data from supranational nonofficial sources should demonstrate whether the collection method has no statistically significant effect on the empirical results.

Screening for PTSD

Screening for PTSD is important, particularly because PTSD is prevalent and costly, but it is not possible to ask more than a few brief questions (Ehring, Klein, Clark, Foa, & Ehlers, 2007) in most settings in which PTSD could be identified (such as pediatric or primary health care, schools, or workplaces). Brief screening measures have been developed to identify persons with or at risk for PTSD and have been tested in psychiatric, medical, community, and school settings (Box 6.7).

Box 6.7

Case Study: How PTSD Assessment Addresses Risks and Informs Treatment

This is a fictional case study based on some of the author’s patients who have been assessed for PTSD. It illustrates the procedures and precautions necessary for PTSD assessment.

Mr. Williams is a 54-year-old African American man who sought treatment for panic attacks and depression at a Medical School community outpatient psychiatry clinic. He is a senior engineer working in the aerospace industry. He been married for 27 years, and he has an adult daughter and an adult son who are married and have children of their own. Mr. Williams described his sister and mother as having needed outpatient treatment for depression and a paternal grandfather who was a problem drinker. He had no previous mental health treatment, was medically healthy, and had no history of head injury, serious past medical illness, suicidality, or substance use problems.

Mr. Williams was making a deposit at his bank 2 months ago when the bank was robbed. Specifically, a noticeably agitated man walked into the bank with a gun and instructed the teller to give him $50,000. Mr. Williams attempted to reason with the robber and keep other customers calm, but the robber became upset and fired his gun, narrowly missing Mr. Williams but hitting another customer. The customer later died from the wounds. After shooting the customer, the armed man had quickly fled the scene and was never apprehended.

Mr. Williams spent 2 weeks on paid administrative leave, which he was instructed to do by his employer. He then returned to work for 4 weeks. Since the attempted robbery, he reported being depressed, feeling jumpy and tense all the time, having serious difficulty sleeping and concentrating, drinking 10–12 beers daily to cope, and going out on sick leave for the past 2 weeks. His wife urged him to get help because his symptoms were worsening. He reluctantly set up an appointment with a clinical psychologist who specialized in the treatment of PTSD.

At his first appointment, Mr. Williams was interviewed in depth about his current and past relationships, personal abilities and problems, educational and work experiences, illnesses, family history of illness, medical or mental treatments, and legal problems or other involvement with the legal system. This information was important in order for the therapist to understand Mr. Williams’s overall life situation before evaluating the possible relationship between the recent traumatic event and his current feelings of “depression.” He was asked these questions because there was a likely possibility of legal involvement, as the bank was being sued by the customer’s family, and because of a possible worker’s compensation case for his emotional condition. The psychologist gathered a detailed demographic, educational, employment, social, legal, family, mental health, and medical history.

In that “intake” interview, the TESI-A was used to carefully explore Mr. Williams’s recollection of and reactions to the bank robbery, as well as to identify any other traumatic stressors that he had experienced before or since the event. The TESI was explained as questions that would help the evaluator and Mr. Williams to efficiently determine if other stressful experiences had happened to him that might affect how he was reacting to the bank robbery. The evaluator was careful to clarify that having prior stressful events would not mean that the robbery had not caused him distress or that his current emotional problems were “preexisting” disorders. He explained to Mr. Williams that when people experience a series of highly stressful or traumatic events, each new one may have a greater effect because of the buildup of stress reactions from the previous ones and that this was the body’s normal, and even healthy, attempt to protect against harm by getting “mobilized for survival.” The psychologist compared this to an airplane’s wings that are always under stress and must have their structural integrity routinely assessed. By using a metaphor that Mr. Williams could relate to, the evaluator was able to reassure him that the inquiry about his trauma history (and next about any PTSD symptoms) was to diagnose fully the source of his current problems so that the best treatment could be provided, and not to point out any failings on his part or to minimize or discredit the distress he’d been feeling since the robbery.

On the TESI, Mr. Williams described having witnessed his father occasionally hit his mother as a child and of having been severely corporally punished (with a bat) and hit in fits of rage by his father. These incidents involved sufficient force that he and his mother had been knocked to the ground and in some instances had suffered broken bones. He also disclosed that he was afraid that his father would lose control so completely that he could kill one of them, even though this never actually happened. This evidence of physical harm and potential life threat meets PTSD’s Criterion A1 for stressors that are objectively traumatic. Mr. Williams also described feeling terrified and helpless most of the times that his father was violent, although he said that he got to the point where he “just didn’t care anymore if he kills one of us or not,” and later felt “so angry that I grabbed the bat out of my father’s hands the last time he tried to hurt her, and by then I was big enough so that he couldn’t overpower me. I didn’t feel scared anymore after that, but I had to be careful not to lose control, or I could have killed my father when he started to make threats.” Mr. Williams thus might appear not to have qualified for PTSD’s Criterion A2—subjective feelings of extreme fear, helplessness, or horror—in the last potentially traumatic violent incident with his father, but he clearly experienced subjective traumatic stress in most of the incidents.

Although he had been in motor vehicle accidents and some fistfights as a youth, and he had been scheduled to go to the World Trade Center on business the week after September 11, 2001, Mr. Williams did not describe any other events that met DSM-IV-TR criteria for traumatic stressors. When asked, he said that the robbery was more troubling to him currently than his father’s abuse and domestic violence, but when memories of the robbery came to mind, he felt as if the robber “changed” and became his father, and then he’d feel almost as frightened as he did when he was a little boy. Mr. Williams was able to recall feeling terrified at first (“When I saw the gun and the crazy look in his eyes, I thought for sure he was going to kill somebody”) but then felt unemotional detachment (“Like I was completely calm, and time slowed down so that everything happened in slow motion, and I could step outside my body and watch like it was just a movie”). Upon inquiry, he recalled having similar detached and unreal feelings that were actually quite pleasant in contrast to the earlier feeling of terror as a child when his father was violent. As he had many times after his father’s violence, Mr. Williams said that when the robbery was over and he knew he wasn’t going to be killed, he had to run to the restroom to vomit. These reactions qualify for PTSD’s Criterion A2—subjective terror and horror—and include reactions probably indicative of dissociation. The veracity of Mr. Williams’s account of the robbery was easily confirmed by checking newspaper reports of the incident.

The evaluator next used the CAPS to assess Mr. Williams’s PTSD symptoms, focusing on his reactions to the recent attempted robbery as well as the additional intensification of these symptoms related to his earlier experiences of abuse and domestic violence. Mr. Williams endorsed three reexperiencing symptoms (intrusive thoughts, nightmares, and flashbacks of the bank robbery as if it was happening all over again; Criterion B), four avoidance and numbing symptoms (persistent avoidance of both thinking about and doing things or going places reminiscent of the robbery, anhedonia (loss of interest in formerly important activities such as his work and recreational activities), and social detachment (feeling emotionally distant from his wife); Criterion C), and three hyperarousal symptoms (difficulty sleeping and concentrating, and hypervigilance; Criterion D). Most symptoms were endorsed as occurring at least several times per week and were sufficiently intense to prevent him from being able to work or socialize. He reported that the symptoms had lasted for nearly 2 months (Criterion E) and that they clearly had disrupted his ability to work (Criterion F). Based on these results, a diagnosis of PTSD, Acute (less than 6 months in duration), was made. Although Mr. Williams described ongoing symptoms of dissociation similar to the dissociative reactions he experienced in the abuse and robbery incidents, he could not be diagnosed with Acute Stress Disorder because the most recent and currently primary traumatic stressor occurred more than 4 weeks before.

Based on the family history and the possibility that Mr. Williams had previously suffered or was currently suffering other psychiatric disorders, particularly depression, dysthymia, or anxiety disorders other than PTSD, Dr. Miller administered the MINI to explore comorbid psychopathology. Based on this interview, Mr. Williams additionally met diagnostic criteria for major depressive disorder, single episode, moderate; and alcohol dependence, without physiological dependence. There was no evidence of any previous occurrence of those disorders, which subsequently was corroborated in an interview with Mrs. Williams (with Mr. Williams’s consent and approval). The evaluator asked Mr. Williams to complete two brief questionnaires assessing depression and alcohol use problems, which indicated that he was not suicidal (corroborating the intake interview’s intensive questions about suicide, homicide, or self-injury risk as not currently present) and that he was experiencing primarily anxiety-related dysphoria (not a more persistent “vegetative” depression) and using alcohol excessively primarily to cope with these feelings and an inability to sleep.

Finally, because of the forensic implications of this case, the evaluator administered the Structured Interview of Reported Symptoms (SIRS) and MMPI-2. SIRS results demonstrated no significant indication of symptom overreporting. MMPI-2 results demonstrated a clinical profile that was consistent with how Mr. Williams described himself in the interviews and questionnaires. Although the MMPI-2 validity scale scores were outside the normal range (L = 40, F = 85, K = 38), the main potential threat to valid self-reporting was an elevated F score that was not more extreme than has been reported in numerous clinical research studies with people diagnosed with PTSD. An F score higher than normal (>65/70) but within three standard deviations of average (<90) has been shown to be suggestive of the extreme anxiety and distress accompanying moderate to severe PTSD.

Based on this assessment, the evaluator proposed a treatment plan beginning with education about PTSD and emotion regulation skills using the TARGET intervention in order to help Mr. Williams gain a greater sense of control over and acceptance of his stress reactions and to develop practical skills for managing stress reactions. Treatment also would use prolonged exposure and cognitive restructuring (PE/CR) to address his immediate traumatic stress reactions related to the recent robbery incident. TARGET and PE/CR also were planned to help Mr. Williams understand his acute dysphoria as an expectable “alarm reaction” and to provide him with skills for improving his mood. Referral to a psychiatrist for an evaluation of the possible benefits of medications that have been found to reduce anxiety and depression symptoms with many persons suffering from PTSD also was discussed with Mr. Williams. After meeting with the psychiatrist, Mr. Williams decided to take an antidepressant medication for a “trial period” of 3 months. With the psychiatrist’s guidance, he eventually discontinued the medication because he was no longer experiencing greater than mild and manageable anxiety or depression symptoms.

Breslau, Peterson, Kessler, and Schultz (1999) developed a seven-item DSM-IV PTSD screen based on results from computer-assisted telephone interviews derived from the National Institute of Mental Health Diagnostic Interview Schedule for DSM-IV and the World Health Organization Composite International Diagnostic Interview, version 2.1. In a community sample (n = 2181) between ages 18 and 45 years old, best-subset regression analysis and receiver operator characteristic analyses were used to select a subset that most efficiently predicted PTSD diagnoses. The final screen includes five avoidance and numbing items and two hyperarousal items. A score of 4 or greater showed it was able to identify adults with undiagnosed PTSD with more than 80% accuracy. Lang and Stein (2005) selected the two intrusive reexperiencing, two avoidance, and two hyperarousal symptoms from the PCL (Weathers, 2008) that best correlated with the total score for its respective PTSD symptom cluster, for a PTSD screen with primary care patients. The PCL screen was more than 80% accurate in identifying patients with undiagnosed PTSD.

Brewin et al. (2002) developed the 10-item Trauma Screening Questionnaire (TSQ) by using the five intrusive reexperiencing and five hyperarousal symptom items from a DSM-IV PTSD symptom questionnaire, modified to be answered as “yes” or “no” within the past week. With samples of adults who had experienced a train crash or crime victimization within the past month (on average, 21 days after the index trauma) and a second sample of assault victims identified in emergency department settings and screened 1–3 weeks later, the TSQ was more than 90% accurate in identifying PTSD cases. Carlson (2001) developed the Screen for Posttraumatic Stress Symptoms (SPTSS) as a brief self-report screening instrument for PTSD symptoms using a low reading level and without requiring a specific traumatic event. In a sample of adult psychiatric inpatients, the SPTSS was more than 80% accurate in identifying patients with undiagnosed PTSD. Zimmerman and Mattia (2001) developed a 15-item PTSD screen (PDSQ-P) for their 122-item Psychiatric Diagnostic Screening Questionnaire, and the screen was more than 80% accurate in identifying patients with undiagnosed PTSD.

The briefest PTSD screening questionnaires are two four-item questionnaires. The DSM-IV PTSD screen for primary care settings (PC-PTSD; Prins et al., 2003) asks a single question for each of the four factor-analytically derived categories of PTSD symptoms: intrusive reexperiencing, avoidance of trauma reminders, emotional numbing, and hyperarousal. The PC-PTSD was more than 80% accurate in identifying patients with undiagnosed PTSD with primary care (Prins et al., 2003) and addictions treatment (Kimerling, Trafton, & Ngyuen, 2006) patients, and incarcerated men and women (Ford, Trestman, Hogan, & Zhang, 2007; Ford, Trestman, Wiesbrock, & Zhang, 2009). The PC-PTSD is available from the National Center for PTSD (www.ptsd.va.gov), which also is leading the development of an updated version of the screen for DSM-5 PTSD.

2.1 Modes of implementation: CATI, SMS, IVR

In phone surveys, there are three main survey modes: Computer Assisted Telephone Interviews (CATI), Interactive Voice Response (IVR), and text message-based surveys (SMS). In CATI surveys, interviewers call respondents to collect data, often from a call center, entering responses into the interactive survey questionnaire on a computer. IVR surveys also contact respondents through voice calls but rely on automated, prerecorded questions instead of live interviewers. Respondents answer to the prerecorded questions by voice or by pressing numbers on the keypad. SMS surveys consist of questions being sent to respondents by text message which they answer by text message prompting follow up questions in an automated fashion (Lau et al., 2019). CATI surveys tend to be more expensive than IVR or SMS but are considered to yield richer information and better data quality (Glazerman et al., 2020). SMS surveys may not be appropriate when large sections of the target population are illiterate. SMS and IVR surveys generally allow shorter, simpler questionnaires relative to CATI surveys. The mode of the LSMS-supported HFPS is CATI, as was the case in many other COVID-19 monitoring surveys.

2.7 Track population-level outcomes

Assessing the impact of public health interventions requires a focus on the wellbeing of entire populations of children and parents. This requires some form of population-level auditing of parents to assess whether parental concerns about children's behavioural and emotional problems have decreased and whether there has been an increase in parent use of positive parenting methods and a decrease in dysfunctional parenting practices. Participation rates in parenting programs and access to formal and informal support should increase.

Previous population trials may inform the selection of measures to be used in trials involving parents of children with developmental disabilities. Sanders and colleagues have used population-level household surveys collected through computer assisted telephone interviews, which have included assessment of constructs that provide population indices of penetration and impact, assessment of practitioners, and evaluation of cost considerations to assess the impact of Triple P (e.g., Sanders, Ralph, et al., 2008). Prinz et al. (2009) employed aggregate archival data at a county level to evaluate the impact of Triple P as a population-level intervention to prevent child maltreatment. The data came from records of statutory authorities that assessed founded and unfounded cases of child maltreatment, out-of-home placements, and hospitalisation accident and injury data. Children with developmental disabilities are at significantly greater risk of being maltreated and receiving an out-of-home placement than their typically developing peers (Sullivan & Knutson, 2000). Consequently, it might be appropriate to use the same variables to assess the impact of a parenting intervention for this population. Other relevant indexes would include measures of childhood development that are routinely and universally collected, such as the Australian Early Development Index (Sayers et al., 2007). These data could be gathered by linking the records of individual children across disability, health and welfare agencies. Such methodology has been successfully used in studies investigating the prevalence and epidemiological characteristics of children with disabilities (e.g., Sullivan & Knutson, 2000).

Implementation

The patient-reported outcome version of the Common Terminology Criteria of Adverse Events19–24 was used most often, followed by the European Organisation for Research and Treatment of Cancer Quality of Life Questionnaire modules QLQ-C30 and QLQ-B23,25–34 the M. D. Anderson Symptom Inventory, the EuroQol-5D questionnaire,24,27,35,36 the Functional Assessment of Cancer Therapy questionnaires,31,33,37,38 and the Hospital Anxiety and Depression Scale,26,30,39 respectively.

PROMs were most often collected electronically (n = 17). Web-based assessments were the primary method of data collection (n = 7), both in and outside the clinic, followed by the use of a tablet (n = 6), an (e-Health) application (n = 2), email (n = 1), and a software system (n = 1) (Table 1). Four studies22,31,33,40 used both electronic and paper-based interventions to collect PROMs. A telephone-based intervention was used in 9 studies, with 4 studies involving interactive voice response,32,41–43 3 studies involving mobile applications,20,44,45 and 2 studies involving semi-structured (computer-assisted) telephone interviews or structured prompts of PRO domains.26,28 One study used only a paper-based intervention.27 Two qualitative studies involving patient focus groups did not include a PROM intervention.46,47 One study48 did not specify the way PRO scores were collected.

The frequency of PROM administration varied from once to daily during the various study periods (Table 1). The duration of PROM interventions ranged from 1.5 months20 to 24 months.48 Most interventions (n = 14) provided PROMs only in the clinic, while 8 studies required patients to complete PROMs at home (all electronic- or telephone-based interventions), and 3 studies allowed patients to complete PROMs either at home or the clinic. PROMs were collected during treatment (n = 22), during follow-up (n = 3), or during both (n = 9). One third of PROM interventions restricted assessments to monitor only a specific phase of breast cancer treatment, most commonly during chemotherapy. The most common methods for reminders were email or phone.

In 4 studies,11,21,29,43 the electronic PRO systems were directly integrated into the electronic health record. Some PROM interventions also provided patient education (n = 9); one of these35 was administered through a module. Of 10 studies11,19,22,23,25,26,28,29,46,47 describing healthcare provider involvement (ie, focus group, interviews, satisfaction or usability questionnaires), only 2 studies19,22 described providers actually being part of the PROM intervention. Within the PROM intervention, patients' providers could edit automatically generated care plans (based on PROM responses) to further tailor referrals and treatment recommendations. Once approved by providers, care plans were mailed to patients.19,22 Outcomes were discussed by the treating clinician (n = 9), nurse (practitioner) or physician assistant (n = 5), or by a combination of these (n = 7) (Table 1). Nearly one third of PROM interventions included alerts being sent to clinicians or patients. Alerts were typically sent by email or text message. In 18 studies, summary reports of PRO data were sent to prespecified providers.

Discussion

Our review identified a mean score for long-term QoL among breast cancer survivors eligible for mammographic screening and followed for >1–10 years since diagnosis to vary from 63.0 (on a scale of 0–100) to 110.5 (on a scale of 0–123).37,38 The studies showed better,37,38 similar,35,36,40 or poorer21,39 QoL among breast cancer survivors compared with women with no history of breast cancer (the reference group). The effect size model based on six studies using a scale from 0 to 100 to measure the mean QoL did not show any statistically significant differences between breast cancer survivors and women with no history of breast cancer, eligible for mammographic screening.

The better results for breast cancer survivors could be explained by the study settings, implying that most of the women had early-stage breast cancer diagnosed in screening programmes.37,38 Furthermore, women attending screening might be healthier and have a higher breast awareness than non-attendees.48–51 On the other side, women usually consider screening as a check and might thus not be prepared for a diagnosis of breast cancer in contrast to women seeking mammography due to symptoms. However, the better results might also have been associated with a relatively short follow-up (1.5–5 years since diagnosis).37,38 Furthermore, no matching by age with the reference group and the possibility of various chronic diseases in the reference group might have resulted in higher scores for QoL and the functioning components for the breast cancer survivors compared with the reference group in one of the studies.37,52

Similar results for breast cancer survivors and women with no history of breast cancer were found in three studies.35,36,40 In one of these studies, the length of the follow-up might have been too short to show any differences (≥14 months since diagnosis).35 In the other study, the data from the reference group were obtained for a long time before the study start, which might have limited health perceptions, as different health awareness, treatments and methods of care were present in that period compared with the study period.36 Furthermore, a small number of women (n = 75) in each group and the length of follow-up of 1.5–3 years might have contributed to the lack of differences. In the other study, the differences in the disease-specific symptom burden implied a less favourable pattern for breast cancer survivors compared with the reference group.40 However, a small number of women (n = 76) with breast cancer vs a large number of women in the reference group (n = 968), and using a 10-year follow-up with a study period 1996–1998, when the treatment recommendations differed from those used in 2000s, might have led to the lack of differences in QoL scores between the groups.40,41 The similar results on QoL, but clinically relevant deterioration in symptoms and several QoL components, could be explained by the response shift or the adaptive mechanisms influencing the overall QoL perception, but not functioning or symptom burden.40,53

The poorer QoL among breast cancer survivors compared with that of women with no history of breast cancer was expected.21,39 However, such results were shown in the study from the United States, including women aged ≥40 years, where younger women were known to have more advanced breast cancer compared with older women.21 The main reasons for this are more aggressive treatment associated with high proliferative aggressive tumours and a stronger impact of treatment on the everyday life of women aged <50 years.54,55 On the other side, screening of women in their 40s is more common in the United States compared with Europe and might mirror the insurance coverage.56 The poorer results of the QoL of breast cancer survivors eligible for screening compared with the reference group in a large population-based study with a 10-year follow-up might be considered the most relevant result of this review,39 assuming that the majority of women in the study population were in the age group 54–75 years. However, the study started in 1995 and was associated with more aggressive treatment for women with early-stage breast cancer and might have impacted the lower scores of QoL among breast cancer survivors compared with women with no history of breast cancer.26,30

Various study settings and periods, numbers of women and lengths of follow-up might have contributed to the results of the meta-analysis, showing no difference between breast cancer survivors and women with no history of breast cancer, eligible for screening.21,35,36,38–40

2 Methods