| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
a MEDTAP® International, Inc., Bethesda, MD,
b Eli Lilly and Company, Indianapolis, IN
c RW Johnson Pharmaceutical Research Institute, Raritan, NJ
Correspondence: Lori Frank, PhD, MEDTAP International, Inc., 7101 Wisconsin Ave., Suite 600, Bethesda, MD 20814. E-mail: Frank{at}medtap.com.
Decision Editor: Laurence G. Branch, PhD
 |
Abstract |
|---|
 TOP Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
Key Words: Questionnaire feasibility • Constipation • Frail elderly persons • Think-aloud approach • Cognitive debriefing
|
Introduction |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
; O'Keefe, Talley, Zinsmeister, and Jacobsen 1995; Schaefer and Cheskin 1998). Definitions of constipation vary widely, and therefore true prevalence estimates are difficult to compare across studies (Lederle 1995). However, there is consensus that prevalence of constipation increases with age; up to 40% of individuals aged 65 or older report constipation problems (Stewart, Moore, Marks, and Hale 1992; Talley, Fleming, Evans, O'Keefe, et al. 1996). Among frail community-dwelling older adults, 45% mentioned constipation as a concern and 11% considered it a major health problem (Wolfsen and Barker 1993). Prevalence of constipation is even higher among long-term care residents: up to 50% by some estimates (Alessi and Henderson 1988; Harari, Gurwitz, Avorn, Choodnovskiy, and Minaker 1994; Williams and DiPalma 1990) and more than 80% by others (Donald, Smith, Cruikshank, Elton, and Stoddart 1985; Read, Celik, and Katsinelos 1995).
Causes of constipation for older adults may include medications, reduced caloric intake, lack of mobility, and ignoring the urge to defecate due to inconvenience or incapacity (Alessi and Henderson 1988; Donald et al. 1985; Read et al. 1995; Romero, Evans, Fleming, and Phillips 1996; Talley et al. 1996; Towers et al. 1994). Treatments generally involve laxatives (Petticrew, Watt, and Brand 1999). Increased fluid and fiber intake and increased activity may also be recommended (Read et al. 1995).
Most current clinical definitions of constipation are based on a measure of frequency of defecation, alone or in combination with other symptoms such as discomfort or straining to defecate (Schaefer and Cheskin 1998). Stool frequency alone is not generally used for diagnosis because of its poor correlation with objective measures (Ashraf, Park, Lof, and Quigley 1996; Koch, Voderholzer, Klauser, and Muller-Lissner 1997). In their review of the literature on constipation in elderly persons, Harari, Gurwitz, and Minaker 1993 found that subjective complaints of constipation increase with age, as does laxative use, but clinical constipation may not. Among community-dwelling older adults, constipation symptoms are common, although bowel frequency may be similar to younger adults; radiographically diagnosed constipation was found in fewer than half of those with symptoms (Donald et al. 1985). For older adults, Harari, Gurwitz, Avorn, Bohn, and Minaker 1997 found that straining and hard stools correlated with self-report of constipation; frequency of less than two per week was not strongly associated with self-report of constipation. High interindividual variation complicates diagnosis by objective criteria (Floch and Wald 1994).
To assist in diagnosis, the Rome criteria (Thompson, Creed, Drossman, Heaton, and Mazzacca 1992) were developed to provide clinicians with guidelines for the diagnosis of constipation. The Rome criteria specify constipation as fewer than two stools per week or two or more of the following: straining with at least 25% of bowel movements, feeling of incomplete evacuation after at least 25% of bowel movements, or hard stools at least 25% of the time. Although it is possible to obtain clinically objective measures such as colon transit time, creation and adoption of the Rome criteria are notable for their focus on patient symptoms. Multiple studies have documented the differences between patients and clinicians in terms of defining constipation (e.g., Herz et al. 1996).
Increasingly, the patient experience of symptoms and condition-specific quality of life are recognized as unique sources of input that can enhance measurement of disorder effects and treatment outcomes (e.g., Moyer and Fendrick 1998). Several clinician rating scales exist (e.g., the Constipation Scoring System [Agachan, Chen, Pfeifer, Reissman, and Wexner 1996] and the Constipation Assessment Scale Constipation Assessment Scale [McMillan and Williams 1989). O'Keefe, Talley, Tangelos, and Zinsmeister 1992 developed the Bowel Symptom Questionnaire for the Elderly, and Osterberg, Grag, Karlbom, and Pahlman 1996 also developed a questionnaire, but neither of these is specific to constipation alone. The Patient Assessment of Constipation (PAC) was developed to provide a brief, easily administered measurement tool for symptoms and QOL aspects of constipation based on the patient's experience of the disorder. The PAC measure was designed to be able to capture both baseline disorder status and response to treatment. It was developed as an outcomes instrument suitable for evaluative research and assessment and designed to be valid and reliable for detecting change in clinical condition. On the basis of earlier validation studies with younger adults (Frank et al. 1999), we have evidence that PAC score change is consistently and meaningfully related to clinical condition change. Because of the rigorous instrument development work, the PAC can also be used in research settings to measure constipation symptoms and clinical course based on patient self-report. It consists of two separate scales, the PAC-SYM, a 12-item measure of symptom severity, and the PAC-QOL, a 28-item measure of the health-related quality of life associated with constipation. They were developed in tandem, and both demonstrate excellent psychometric properties in use among community-dwelling younger adults (Dubois, Johnson, de la Loge, and Marquis 1998; Frank, Kleinman, Farup, Taylor, and Miner 1999; Marquis, Dubois, Frank, and Rothman 2000). However, because these studies were conducted among community-dwelling adults generally under the age of 65, the usefulness of the PAC for older patient populations is unknown. A patient-based constipation assessment tool applicable to older adults in long-term care would be particularly valuable given the prevalence and serious sequelae of constipation among that population.
The quality of questionnaire data obtained from older adults, particularly institutionalized older adults, may differ from that obtained from younger adults. For example, older adults may interpret items differently or have difficulty with memory involved in retrieving relevant information (Jobe and Mingay 1990; McHorney, Ware, Lu, and Sherbourne 1994). Also, West, Bondy, and Hutchinson 1991 have pointed out that physical, cognitive, affective, and personal factors may threaten the validity of interview data from institutionalized older adults. These findings have consequences for accurate and valid symptom and QOL measurement among institutionalized older adult populations, given that data quality is fundamental to establishing the reliability and validity of an instrument. The potential differences between younger and older adults in response to the same questionnaires create the need to closely examine the use of questionnaires with older adults, particularly for those developed and tested among younger adults as was the PAC.
In addition, gerontology researchers have long recognized the unique concerns of older adults in nursing homes with regard to questionnaire administration. Accurate communication of questions and response choices is one key concern. To accommodate visual impairments among some subjects, font size of written materials is often increased. The mode of administration is another consideration. Some evidence suggests questionnaire items can be more successfully communicated to older adults through interviewer-administered rather than self-administered questionnaires (e.g., Coast, Peters, Richards, and Gunnell 1998).
Understanding the ways in which subjects respond to items is a key step in evaluating a questionnaire. Thorough knowledge of the subjects' thought processes involved in responding ensures that the items are performing in the manner intended. A cognitive debriefing was included as a step in instrument development for the PAC, but because the PAC has not yet been used with older adults, no information is available about the older adults' thought processes in response to PAC items. For this study, the cognitive process model (Jobe and Mingay 1990; Warnecke et al. 1997) was used as a framework for interpretation of results from the cognitive debriefing. According to cognitive process theory, subjects perform four tasks when responding to a question: interpretation, memory retrieval, judgment formation, and response editing (Warnecke et al. 1997). In this way, we address the potential for "category fallacy," the incorrect assumption that item interpretation is the same for the group being tested as for the group from which the question was developed (Warnecke et al. 1997).
The objective of this study was to evaluate the feasibility and validity of the interview format for administration of the PAC to older adults residing in nursing homes. Specifically, the following attributes were examined: acceptability of items to subjects, acceptability of time to complete the questionnaire, extent of missing data, and ability of subjects to understand and answer the items (Coast et al. 1998; McHorney et al. 1994). Cognitive interviewing methodology, a set of qualitative research methods developed to improve questionnaires and surveys, was used to address the objective. Both the think-aloud approach and cognitive debriefing are part of cognitive interviewing methodology. Researchers involved in questionnaire design have long recognized that the wording of questions, the format of responses, and other aspects of questionnaires can be readily misunderstood or misinterpreted by some subjects, leading to inaccurate measurement. To help avoid these mistakes, subjects are asked to verbalize their thoughts (think aloud) as they answer the questionnaire. Cognitive debriefing involves probing techniques, in which researchers can ask subjects to further explain certain responses, either during administration of the questionnaire or following completion of it (Jobe and Mingay 1989; Willis 1994). These methods help to identify problems with the questionnaire that could limit the validity or reliability of subjects' responses.
|
Methods |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
) and in an effort to minimize respondent burden, only this subscale from the PAC-QOL was included. The entire 12-item PAC-SYM was used. This study was approved by one central and four site-specific institutional review boards. The interviewer explained the study to eligible subjects and obtained written informed consent from subjects before beginning the interview. For subjects with legal guardians or for whom cognitive capacity was in question, the interviewer contacted the guardian or appropriate family member, provided copies of the informed consent form, and obtained written informed consent.
Subjects were required to be 65 years of age or older, to be clinically stable, to be able to take oral medications and to use a toilet or bedside commode, to have the ability to reliably communicate symptoms as judged by nursing home staff, and to have a history of constipation, defined as use of a stimulant or osmotic laxative or enema at least once a week for the 4 weeks before the interview. Nursing home staff were asked to assist with the identification of individuals who met the eligibility criteria. The interviewer obtained background clinical and demographic data from the medical record and/or from nursing staff.
This study was conducted in two phases. The first phase was a pilot study to test and refine the interview methods. Five older adult subjects in long-term care were interviewed, using the PAC. For Phase 1, the recruiting nurse was asked to identify subjects likely to have the cognitive capacity to respond to a face-to-face interview about constipation.
An interviewer read the PAC items while subjects looked at a large-font response option card during the interview. Subjects were asked to think aloud during the actual PAC interview, to obtain a record of their thinking during the process. At the completion of PAC administration, subjects were asked to respond to semistructured cognitive debriefing questions.
The debriefing questions for this study were drawn from the cognitive debriefing process used in the development of the PAC-SYM and included special attention to subject interpretation of the Likert scale anchors and to specific item wording that was a concern in subject interpretation during PAC development. Debriefing questions were organized into the following sections: overall impressions, general instructions and recall period, items, meaning of item terms, and meaning of item anchors.
On the basis of the pilot study, the interviewer was instructed to reorient the subject to the response options by rereading the Likert anchor points on the response option card for each item rather than allowing the subject to read the response option card in silence. The interviewer also reread the stem question for each item because the pilot study also indicated that this improved subject item understanding.
In Phase 2, a total of 19 subjects were interviewed. Subjects were recruited from five nursing homes located in the northeastern United States. Three sites were nonprofit nursing homes of between 100 and 340 beds; two sites were for-profit nursing homes of between 175 and 300 beds. Because Phase 2 was the formal extension of the initial work and involved a larger sample and because multiple nurses were assisting with recruitment, assessment of cognitive status was formalized for Phase 2. Extent of cognitive impairment was rated by nursing staff or the interviewer, using a 5-point scale from none to severe. Subjects with cognitive impairment rated as none or minimal were selected for Phase 2.
Data on some personal characteristics were also collected (race, education, prior occupation, and ethnic background). Diagnoses and patient self-rating of constipation severity (on a 7-point scale) were collected to understand the relationship of responses to physical factors. All Phase 2 interviews were conducted by a single interviewer.
The interviewer read the PAC instructions to subjects and then asked the specified PAC questions, PAC-SYM first, PAC-QOL second. Acceptability of items, acceptability of time to complete, ability of subjects to understand the items, and questionnaire face validity (subject perceived relevance) were all examined through cognitive debriefing, as described above. During the PAC administration, subjects were asked to verbalize their thoughts as they completed each item, and their comments were recorded. At the completion of the PAC interview, subjects were asked semistructured questions about the questionnaire administration. The time to complete each PAC interview was measured, although those times are likely to be longer than for usual administration since they included time to "think aloud."
All interviews took place in private, in the subject's room or in a quiet office or conference room. For all interviews, the lighting was good. At times there was slight background noise from other residents and staff. At no time was this disruptive to the interview.
|
Results |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
|
Quantitative Results
Frequencies of item responses are presented in Table 2 and Table 3 for the PAC-SYM and PAC-QOL, respectively. There were no missing data on the PAC-SYM. In general, subjects responded across the range of scale choices. Three subjects chose "absent" in response to all 12 PAC-SYM items. Of interest is the high proportion of patients responding "absent" to the items regarding rectal burning (17 of 19) and rectal tearing or bleeding (16 of 19).
|
|
Cognitive Debriefing and Think-Aloud Results
Responses to the cognitive debriefing questions, asked after administration of the PAC, are summarized below.
Overall Impressions.
The first questions of the cognitive debriefing interview concerned content validity and comprehensiveness, ease of administration, and interview length. All 19 subjects reported that the interview asked the right questions (addressing relevance as perceived by the subject, or face validity). Three subjects had suggestions on other areas to ask about. One suggested asking "about blood more," one suggested asking about diet, and 1 subject suggested asking about laxatives.
One subject mentioned that we did ask if subjects had to change their routine in response to constipation. (The daily routine was important to this subject, and a change in bowel habits caused a change in routine for this 1 subject). The interview was "moderately short" to 1 subject and the "questions simple." Another subject called the interview "very sensible." Just 1 subject noted it was difficult to complete and long. This same subject responded negatively in general to the entire interview, calling it "embarrassing."
General Instructions and Recall Period.
The next questions in the cognitive debriefing interview addressed whether the instructions were clear, if the subject limited recall to the prior week only, and if recall was difficult. The instructions were clear to the subjects: "nothing confusing," "fine," "clear as day." One subject said that the language was not in medical terms, which was good because she does not understand medical terms.
In response to the question "Did you think about the past week only?" 7 of the 19 subjects said no. One of those 7 noted that she thought instead about "the whole thing, of course. What good is the past week?" Another noted that she "has the problems every day." Another said she "thought about all the time," and another replied she thought about "more than the past week." Two other subjects who said they did limit their thoughts to the past week (as instructed) had interesting comments in response to that question. One said, "Yes, [it was] difficult but [I] tried to continue." Another subject remarked that "one week doesn't seem to be enough time, maybe one month." Two other subjects remarked that they thought about one particular day with problems as they responded to the PAC-SYM. To determine if cognitive impairment could account for their different recall period, we examined the clinical and cognitive status variables for these subjects. There appeared to be no relationship among these subjects in terms of their self-rating of constipation severity (ranging from 1 [mild] to 5), their self-reported duration of constipation (ranging from less than 3 months to "all my life"), or their nurse-rated cognitive status (mild to minimal).
Items.
On the basis of results from our pilot testing with older adults in nursing homes for this project, the interviewer reread each response option following the reading of each item as a means of ensuring subject orientation to the response scale for each item. However, subjects would often interrupt the interviewer as she read their preferred response choice. For example, rather than letting her keep reading the response choices of "severe" and "very severe," the subject would reply "moderate" immediately after the interviewer read that as a response option. Per protocol, the interviewer continued to reread the response options anyway. All subjects heard the full set of response options read at least once. Most subjects appeared to understand the response options for the PAC-SYM; fewer subjects understood the response options for the PAC-QOL and thus asked the interviewer for clarification.
The first three cognitive debriefing questions about the items were intended to determine if specific items were difficult to understand, what if any changes they would recommend, and whether they thought other subjects might have difficulty responding to some items. This last question allowed for elicitation of modification suggestions even if the subject him- or herself was unwilling to admit to difficulty with understanding.
No subjects reported difficulty understanding any items themselves. Only 4 of the 19 subjects responded affirmatively to the question "Do you think certain questions will be hard for other people to understand?" (One of these was a subject who reacted negatively to most of the interview.) One noted that it may be "hard for people to remember if they had a problem." Another said, "It might be for some. Some people just don't understand." In general, however, subjects reported that the items were readily understandable. One said the questions were not difficult, it "just takes time to come up with the answer."
Meaning of Item Terms.
The next questions in the cognitive debriefing were meant to elicit the subjects' interpretation of the terms abdomen and stomach, specific terms used in the PAC. Use of these terms in gastrointestinal surveys has come under increased scrutiny, as researchers have noted that consistent interpretation of these terms across subjects is required for accurate interpretation of results from questionnaires in which they are used. Focus group results from the initial development of the PAC suggested that U.S. subjects may interpret these terms in different ways, potentially altering the intended symptom description. Eight subjects explicitly equated abdomen and stomach. Nine referred to the stomach as "where the food goes," "put food in it," "part of digestive tract," and similar responses. One of these subjects pointed to her waist to clarify her meaning for both terms. One subject identified the abdomen as below the stomach. Another noted that the "stomach was part of the abdomen," and another said she was "not sure of the difference." The remaining subjects did not provide a definition of either stomach or abdomen and appeared to not understand the point of the question. In general, subjects more readily offered definitions for stomach than for abdomen. The order in which the questions were asked (abdomen first) may have influenced results. In general, subjects' responses suggested accurate interpretation of the items.
The next question was "Are rectal burning and painful bowel movement the same or different?" These terms were intended to represent distinct symptom sets. Eleven of the 19 subjects responded that they are different, with one noting that "painful bowel movement causes rectal burning." Six said they were the same or the "same classification," and 1 did not understand the point of the question. Another subject said, "I don't think about them" and provided no other answer.
In response to the question "What does the word ‘regularity’ mean to you?" all 19 subjects responded with consistently similar answers (e.g., "going every day" or "daily"). The next question asked, "What does it mean to be satisfied with your treatment?" As with the prior question, the responses were generally consistent: "happy," "satisfied," "moderately happy," "meds are working; comes out without any trouble," "methods take care of problem," "happy, OK with what the doctors do," "they do something that works," "fine," and "if I need something I get it." One subject would not answer and appeared agitated and annoyed with cognitive interview questions 7 through 12. Another subject did not appear to understand and replied "be happy with people, but some people I can't stand." This same subject frequently mentioned her dislike of another resident throughout the interview, even in response to some of the questions, although the other resident was not in the room at any time during the interview.
Meaning of Item Anchors.
Subjects were asked to explain the meaning of the Likert scale anchor terms for mild to very severe. Three of the 19 subjects did not reply to the question at all, 2 of the subjects responded to only the first anchor question, and 1 responded only to the first and last anchor questions. Some expressed irritation at the questions about the meaning of the anchors. Of the 13 subjects who attempted responses to all anchor options, 7 presented explanations that indicated understanding of the ordinal nature of the scale. The other 6 presented answers that suggested increasing severity across options, but equated one or two adjacent response options. These results are similar to those obtained from pilot testing with 5 younger adults during initial development of the PAC-SYM, in which subject responses indicated an appropriate ordinal ranking of these anchors. None of the younger subjects equated response options, however; all were verbally distinguished.
The next question asked subjects to describe the meaning of the PAC-QOL Likert anchors (not at all, a little bit, moderately, quite a bit, or extremely satisfied). As noted above, some of the cognitive debriefing questions asking subjects to define Likert anchor terms were annoying to subjects. Whereas a total of 5 of the 19 subjects did not respond or responded minimally to the same set of questions regarding the PAC-SYM scale anchors, 9 of the 19 subjects did not respond to the questions about the PAC-QOL anchors. Among those who responded, there was general understanding of the specific terms and of the progressive, ordinal nature of the scale.
General Comments.
Finally subjects were asked if they had any other comments on the interview. One said, "They give me enemas; you didn't ask me about enemas," and another said, "Very strange to have someone ask these questions; hard to measure; [I] don't think of constipation and satisfaction" (they don't go together in his mind). One subject No. (14) noted that the "categories were tough." Another subject (No. 15) noted that the questionnaire was "too general" but then said, "I don't know if you can get more specific ... if you don't make it general you get problems too."
|
Discussion |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The relevance of constipation research assessment methods to long-term care residents is underscored by the high prevalence of constipation in the long-term care population as well as the high care costs associated with constipation in long-term care (Frank et al. 2000). Because the PAC has demonstrated reliability, validity, and responsiveness to change over time among community-dwelling adults (Frank et al. 1999), it should be of value to research and assessment for community-dwelling older adults as well. However, the PAC is not intended as a diagnostic tool, and although it might serve that function, more research would be required (i.e., comparison with clinician diagnosis) before its usefulness in diagnosis could be confirmed.
The 1-week recall period may not be appropriate for this population. More than a third of subjects reported that their responses were based on a longer time frame or a more general impression instead. This differs from results from the pilot testing with younger adults, in which all subjects indicated that they limited their responses to the prior week.
Several subjects reported no constipation symptoms based on responses to PAC-SYM items. Subjects were selected on the basis of presence of constipation (frequency and/or use of specific medications), and it is possible that their treatment regimens provided adequate symptom relief accounting for the apparent lack of current symptoms. Given the relatively low level of symptom severity of this sample as noted in the PAC-SYM, a slightly longer recall period may provide more of a symptom experience base from which to respond. The self-administered PAC was originally developed using a 2-week recall period to ensure occurrence of events to rate. Although that recall period was chosen in part to accommodate individuals with severe constipation and bowel movement frequencies of less than 2 per week, the same reasoning may apply to individuals with reasonably well-controlled constipation symptoms. The finding of low symptom severity among this sample is consistent with results from a study of 59 constipated residents in long-term care (Frank et al. 2000). In that study, all residents identified by nursing staff as suffering from constipation, were receiving symptomatic treatment for constipation, and therefore symptoms were minimal or absent for some subjects.
The cognitive debriefing interview was designed to determine if subjects consistently interpreted the meaning of specific terms: stomach versus abdomen, rectal burning versus painful bowel movement, and regularity. Interpretation of stomach and abdomen and description of interpretation of the items regarding abdominal discomfort, pain, and bloating appeared to be consistent, as did interpretation of the item about stomach cramps. The corresponding symptom descriptions appear to have been interpreted as intended by the authors of the scale by these respondents (face valid). Having separate items for rectal burning and painful bowel movements appears justified; nearly two thirds of subjects explicitly stated these were distinct. Subjects readily understood and consistently interpreted the concept of bowel movement regularity.
Consistent interpretation of the Likert response scales was another concern. When asked to explain the meaning of the response options, most subjects' responses indicated understanding of the ordinal nature of the scale. As noted above, most subjects appeared to understand the severity response options for PAC-SYM but found the response options for the satisfaction scale of the PAC-QOL more difficult.
The cognitive process model has been invoked to examine the cognitive processes used by subjects to respond to questionnaire or survey items (Jobe and Mingay 1990; Warnecke et al. 1997). According to cognitive process theory, subjects perform four tasks when responding to a question: interpretation, memory retrieval, judgment formation, and response editing (Warnecke et al. 1997). Concepts of cross-cultural psychology apply to establishing measurement equivalence even across different demographic groups within a culture. To cross-cultural researchers, a category fallacy can result from the incorrect assumption that item interpretation is the same for the group being tested as for the group from which the question was developed (Warnecke et al. 1997).
The cognitive debriefing questions were designed to illuminate the interpretation of questions. Consistent item interpretation is fundamental to the accuracy of data collected and interpretations based on those data. Results from this study suggest that the PAC instructions and most items were accurately interpreted by most subjects, and some form of ordinal scaling was implied by the subjects' responses about the Likert response options. There was less response in general to the question that asked for definitions of the PAC-QOL response options, and some subjects indicated problems with the "extremely" choice (Subject 13 said, "That's ridiculous," and Subject 15 said, "I'm rather dubious about that one"). As noted below, the English version of the response options for the PAC-QOL may need further consideration. Although psychometric performance of the PAC-QOL was evaluated and found acceptable in a U.S. population, some wording problems may remain as it was developed in a non-U.S. population. The PAC-QOL item regarding satisfaction with treatment caused confusion among subjects, in large part due to subject uncertainty about the meaning of the term treatment. In the context of a clinical trial or evaluation of a specific treatment regimen, this item may cause no difficulty, but clarification regarding what is meant by treatment may help ensure consistent interpretation of the item across subjects.
The second task subjects must perform when responding to a question is memory retrieval. Cognitive process theory suggests that regularly occurring events are likely to be remembered as generalized schemas rather than as episodic discrete events. This point has direct relevance for the PAC recall period used with different patient populations. Clearly, this sample of older adults tended toward generalized schema memory, with many readily admitting that they did not limit their responses to the past week as instructed but instead recalled across longer time periods and, presumably, obtained a more general impression on which they based their responses. The length of the recall period for the PAC was originally set at 2 weeks to allow for sufficient occurrences of constipation-related events to permit rating. The "acute" 1-week version may result in more generalized schema recall among respondents, particularly if few constipation-related events occurred in the week prior. If this holds true, then the responsiveness of this instrument to change over time will be attenuated.
The third task is judgment formation. Cognitive process theory suggests that episodic, discrete event recall is less subject to various aspects of cultural bias than general schema recall. If this is true, then encouraging respondents to truly focus on specific events in the prior week could improve data quality by minimizing the potential bias and distortion introduced by cultural influences on response (e.g., avoiding extremes on scales, modesty leading to severity minimization).
The final task is response editing. Social desirability, a recognized threat to validity, may lead subjects to answer in the manner they think is expected of them. One subject even commented when asked if the interview was difficult to complete that she had a "good interviewer." In response to the question about the length of the interview, this same subject said, "Oh, I like talking to you. [The length] is just fine." These responses potentially indicate both the influence of loneliness and the related desire to please others and provide them what the subject thinks they want. The nursing home setting is likely to enhance these editing tendencies.
The greater the response editing, the greater the bias introduced. As Warnecke and colleagues 1997 noted, editing is more likely when there is a cultural difference between the interviewer and respondent. The implication for this study is that the age of the interviewer may have affected the responses obtained. Our interviewer was 31 years old. More closely age-matched interviewers may elicit less editing, although for obtaining patient response to the actual questionnaires this sensitivity to interviewer age is probably not warranted. Further, nursing staff with whom nursing home residents interact frequently are younger than most residents themselves. There is no evidence to suggest that older nursing home residents edit responses to nursing staff regarding clinical issues. Although this is an interesting question for further consideration, accuracy of responses to the interviewer-administered PAC is unlikely to be dependent on interviewer age.
Among the pieces of evidence identified to help with the evaluation of the suitability of an instrument is the extent of missing data (McHorney et al. 1994). It is therefore of interest that only two items had any missing data, PAC-QOL Items 2 and 4. Item 4 reads, "To what extent during the past week have you been satisfied with the time it takes for food to pass through the intestines?" Four subjects did not answer this, with 1 explicitly stating she could not. The interviewer noted, however, that many subjects appeared to have difficulty interpreting Item 4. Further exploration of the meaning of this item to subjects could be helpful.
As noted above, some subjects expressed frustration at the rereading of all response choices with each item and with the repetitive nature of some cognitive debriefing questions. Because these subjects had little or no cognitive impairment, the rereading strategy established on the basis of the pilot study may have been annoying or insulting to them. In actual PAC administration, cognitive debriefing questions are not asked, but sensitivity to subject annoyance is prudent. Informing subjects that the interviewer will be reading each response choice repeatedly may be one way to improve patience with the practice. Alternatively, this practice could be modified depending on interviewer judgment of adequate patient orientation to the response choices, especially as all subjects are given printed copies during the interview.
Keeping subjects on topic is also anecdotally a challenge when interviewing older adults. This may be particularly salient in a nursing home where interaction with other people is limited, making the interviewer a welcome "guest." Administration times for 4 of the subjects reflect this; these subjects became quite conversational during the PAC administration. While practical time considerations may require keeping the subject more tightly on track, allowing sufficient time for some conversation aids in establishing rapport and should be anticipated in an interview format administration with older adults. Discussion during the interview should be discouraged, however, as it can lead to inconsistent administration across subjects and therefore to bias.
Use of an open-ended interview approach to the study of symptom and quality-of-life measures with older adults is supported by evidence from qualitative research (e.g., Lincoln and Guba 1985, and Steckler, McLeroy, Goodman, Bird, and McCormick 1992, as cited in MacEntee 1996). Our results are consistent with those of other researchers, who have demonstrated that the think-aloud and probe question approach are effective means of obtaining information to maximize data quality from older adults (Jobe and Mingay 1990).
The results from this study suggest that the interviewer-administered PAC is feasible for use with institutionalized older adults. This study also suggests several avenues for future work. As results from use of the EuroQoL with older adults have suggested (Coast et al. 1998), comparison of self-administered and interviewer-administered results is essential to understanding the contribution of administration format to results. A direct comparison in this population may therefore yield additional insight about the use of the PAC. As Gregoire, de Leval, Mesters, and Czarka 1994(p. 14) pointed out, questionnaire validation "is an endless process of accumulating evidence." Although the results from this study support the feasibility, acceptability, ease of administration, and intended interpretation of the PAC for this population, quantitative study of the psychometric properties of the PAC in this population, in comparison to the population on which it was first tested, could be illuminating. These qualitative results are the first step in the validation process for the population of older adults in long-term care.
Practice Concepts
The Forum
Book Reviews
Received for publication December 11, 2000. Accepted for publication June 22, 2001.
|
References |
|---|
|
TOPAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
