Development of the thirst distress scale.
Thirst is the most frequently occurring symptom in hemodialysis patients (Virga et al., 1998). Intervening to reduce the symptom of thirst (Welch, 2001), therefore, may provide a meaningful way to decrease IWG and prevent associated complications. Unfortunately, limited empirical work has been done to investigate the symptom of thirst or to develop reliable and valid measures of thirst. This work is needed before interventions can be developed to reduce thirst in these vulnerable patients. The purpose of this study was to develop and test the psychometric properties of an instrument to measure thirst.
Review of Literature
Items for this measure were developed based on an adaptation of the University of California, San Francisco School of Nursing (UCSF) Symptom Management Model (Dodd et al., 2001; UCSF, 1994). In this model, a symptom is defined as a subjective experience "reflecting changes in a person's biopsychosocial function, sensation, or cognition." Symptoms are undetectable by others. Two important components of the model include symptom evaluation and symptom response. Symptom evaluation refers to the judgments people make about their symptoms, such as frequency, duration, intensity, and distress. Symptom response may include physiological, emotional, and/or behavioral expressions of a symptom, such as physical, cognitive, and affective changes. In this study, the conceptual framework (see Figure 1) proposes that a positive relationship exists between thirst distress and, respectively, thirst duration, frequency, and intensity. A positive relationship between thirst distress, duration, frequency, and intensity, and respectively, IWG is also proposed.
[FIGURE 1 OMITTED]
Although thirst is triggered by the hypothalamus in response to changes in body fluids (Porth & Erickson, 1992), definitions of thirst describe it as a symptom. For example, thirst has been defined as a "subjective symptom or sensation associated with a craving to drink" (Porth & Erickson, 1992), a "motivational state in which there is a sensation of a need or desire for water" (Woodtli, 1990), a "sensation of dryness in the mouth and throat associated with a desire for liquids" (Greenleaf, 1992), and as a "deep-seated sensation or desire for water that cannot be ignored and causes a powerful behavioral drive to drink water" (Toto, 1994). Because these definitions suggest that the sensation of thirst is subjective, using self-report methods to measure thirst is appropriate. In this study, the state of being thirsty is defined as a conscious and subjective sensation of desiring fluids.
Persons receiving hemodialysis experience intense thirst (Virga et al., 1998). Moreover, the literature suggests that adults receiving hemodialysis who have more thirst have greater IWG. For example, in one study with 32 hemodialysis patients, Wirth and Folstein (1982) examined relationships between thirst and IWG. Intensity of thirst, as measured by a visual analogue scale, was assessed for "right this minute" and "the past 2 months on dialysis." An increase in thirst over the preceding 2 months, as well as current thirst, was associated with increased IWG. In another study of 32 hemodialysis patients, a 15-item scale assessed perceived thirst and drinking difficulties; no psychometric data were presented on the scale. In this study, an increase in thirst was associated with an increase in both fluid intake and IWG (Oldenburg, Macdonald, & Perkins, 1988). In a third study of 247 hemodialysis patients (Giovannetti et al., 1994), thirst intensity was measured with one item using a 4-point response scale from 0 (no abnormal thirst) to 3 (almost always present). Exaggerated thirst was present in 86% of the participants. Although it was not clear when the thirst assessment was made, an increase in thirst was associated with increased IWG. In a fourth study (Brady, Tucker, Alfino, Tarrant, & Finlayson, 1997), the side effect of thirst from some antihypertensive medications was proposed as a potential cause for poorer fluid adherence. Thirst was not directly measured in this study.
Although few in number, these reports suggest that thirst is an important variable to consider and might be a potential target for nursing interventions to reduce IWG. Despite the severity of thirst experienced by hemodialysis patients, nurses know little about the phenomenon and, unfortunately, no nursing research has been conducted on thirst. Furthermore, instruments to assess thirst are lacking.
According to the Symptom Management Model (UCSF, 1994), a complete evaluation of any symptom experience must include an assessment of distress, duration, frequency, and intensity. Thirst intensity, defined as the severity, strength, or amount of thirst, has been successfully measured in past research using a visual analogue scale (Dominic, Ramachandran, Somiah, Mani, & Dominic, 1996; Martinez-Vea, Garcia, Gaya, Rivera, & Oliver, 1992; Phillips, Bretherton, Johnston, & Gray, 1991). Therefore, only three different dimensions of thirst were conceptualized in this instrument development study: distress, duration, and frequency. Conceptual definitions for each domain were as follows:
1. Thirst distress: the degree to which a person is bothered by thirst or its associated discomfort.
2. Thirst duration: the length of time that thirst is experienced by the person per episode.
3. Thirst frequency: how often during a day thirst is experienced by the person.
Items were initially developed for each of the three domains using established guidelines (DeVellis, 1991; Dillman, 1978; McIver & Carmines, 1981) and were based on interviews with 10 patients, a literature review, and the theoretical definitions.
The use of 3 to 10 experts is recommended as one step in determining content validity (Lynn, 1986). The proposed items were sent to a panel of 9 individuals: 3 doctorally prepared experts in instrument development, 3 doctorally prepared experts in symptom management, and 1 doctorally and 2 masters-prepared nurses in nephrology nursing. The judges were provided with the conceptual definition for each thirst domain and asked to classify each of the items into one domain. Items in each domain were retained if at least 8 of the 9 judges classified the item in the appropriate domain. In addition, experts judged each item for its relevancy to the conceptual definition. All items were assessed on four-point rating scales ranging from 4 (extremely relevant to conceptual definition) to 1 (not relevant). A content validity index greater than .78 from the nine-member panel was needed for item retention (Lynn, 1986). Revisions for item clarity were made following review by the experts. Of the original items, 31 were retained for further testing: 12 addressing distress, 3 duration, and 16 frequency.
Reliability and Validity Testing
Subjects. Subjects in this study included 247 adults age 18 or older receiving thrice-weekly outpatient hemodialysis therapy. A convenience sample was obtained from four hemodialysis units: one urban, one rural, and two suburban. A total of 277 individuals were invited to participate, and 247 (89%) gave informed consent for a short interview. The age range for the sample was 20-91 years, with a mean of 59 years and an SD of 14.8. A majority of the sample (60%) was Caucasian, 36% were African-American, and the remainder (4%) were Asian, Hispanic, or Native American. The mean educational level was 11.8 years, with a range of 4-22 and an SD of 2.8.
Procedure. Research assistants approached adults receiving outpatient hemodialysis during a midweek scheduled treatment. After obtaining verbal informed consent, a research assistant read each item to the participants during the hemodialysis session. Participants were asked to verbally respond to the items based on experiences with thirst since their last dialysis treatment; the interview lasted approximately 15 minutes. The items were scored on 5-point rating scales from 1 (strongly disagree) to 5 (strongly agree).
Other measures. To test for the construct validity of the thirst measure, measures for each of the constructs in the conceptual framework were included in data collection.
IWG. Because participants responded to the thirst items based on experiences since the last dialysis, IWG was computed for only the immediately preceding interdialytic interval. To obtain IWG, the predialysis weight on the day of the interview and the postdialysis weight from the preceding dialysis were obtained. IWG was calculated by subtracting the previous postdialysis weight from the current predialysis weight. The IWG was divided by the number of days between treatments in order to standardize differences in the number of days between treatments. The mean daily IWG was used for analysis. In this sample, individuals' mean daily IWG since the last hemodialysis treatment ranged between -0.5 and 3.63 kg (M = 1.24; SD = .70).
Thirst intensity. Although thirst intensity was not included in scale development, it was included as one method to assess construct validity. Thirst intensity was measured using a 100 mm visual analogue scale (VAS) consisting of a 100 mm line with the degree of their thirstiness quantified on the left as not at all (0 mm) and on the right as extremely (100 mm). Patients were asked to mark the line at the point that best described how they would rate their worst thirst since the last dialysis. Individuals' thirst intensity, as measured by the VAS, ranged between 1 and 100 (M = 54.18; SD = 29.41).
The hypotheses guiding data analysis were as follows:
1. Cronbach's coefficient alpha will be above .7.
2. All items will have a factor loading of .3 or greater on one dimension via factor analysis.
3. Thirst distress will be positively related to IWG.
4. Thirst duration, frequency, and intensity will be positively related to thirst distress.
Data were analyzed using the Statistical Package for Social Sciences (SPSS 10.0) and LISREL (LISREL 8.3). Reliability was assessed using item analysis, and coefficient alpha. Validity was tested using confirmatory factor analysis, exploratory factor analysis, and relationship testing of constructs identified in the theoretical model.
Assessment of reliability. Reliability is defined as the consistency of measurement of the same phenomenon (Carmines & Zeller, 1979). The first step in the psychometric assessment was item analysis for each subscale. Because an average inter-item correlation below .30 indicates that an item is not sufficiently related and does not contribute to the measurement of the latent variable and an inter-item correlation above .70 indicates the item is redundant (Ferketich, 1991), an item was only retained if its average correlation with other items on the scale was .30 or above but below .70. All items in the duration and frequency subscales were deleted because none met these criteria. Analysis of the remaining 12 distress items resulted in the deletion of 6 items because their average inter-item correlations were below .30. The final scale had 6 items (see Table 1).
It is recommended that item means be near the center of the possible range (DeVellis, 1991), and the item means for the six remaining items on the thirst distress scale were near the midpoint. Standard deviations of the items ranged between .95 and 1.04, indicating good variability in relation to the means. A corrected item-to-scale correlation is a summary statistic, which, when combined with other measures, provides a comprehensive view of item performance. Corrected item-total correlations above .30 and less than .70 are recommended (Ferketich, 1991) and the correlations for the 6 items ranged from .43 to .68 with no improvement in alpha if particular items were deleted. The Cronbach's alpha for the thirst distress scale was .78, which is satisfactory (Nunnally, 1978) and provides support for hypothesis 1. All item statistics are shown in Table 2.
Assessment of validity. An assessment of validity determines whether an instrument measures what is intended (Polit & Hungler, 1999). Estimation of validity was determined using three different analyses: confirmatory factor analysis, exploratory factor analysis, and Pearson correlations. A confirmatory factor analysis, which uses a covariance matrix of scale items, was conducted as a measure of construct validity. The analysis indicated that error covariance was present among two sets of two items. Although the inclusion of the error covariance would improve the fit statistics, the model was not modified because it would also increase the difficulty of interpretation. The goodness-of-fit index, an index indicating the amount of variance and covariance accounted for by the model (Long, 1983), was .94, indicating a good fit. The chi-square goodness-of-fit statistic is sometimes used as an additional indication of the data-model fit. This statistic will be small and nonsignificant if there is a good fit between the data and model. Because the chi-square statistic is affected by sample size, Boyd and colleagues (1988) recommend assessing the relative chi-square, defined as the ratio of chi-square to degrees of freedom. The chi-square for this solution was 40.70 and there were 9 degrees of freedom, yielding a relative chi-square of 4.52. Although relative chi-square values less than 3.0 are preferred, values as high as 5.0 represent good data-model fit (Aroian & Norris, 2001). The standardized solution for each item is displayed in Table 3 and the t-values for all items were significant.
Because the relative chi-square was at the high end of acceptable values, the scale was further explored using exploratory factor analysis. In a unidimensional scale such as the thirst distress scale, the summed item scores should measure the attribute under investigation (McIver & Carmines, 1981). Statistically, this means that the thirst distress items were expected to group together because they were measuring one construct. Factor analysis using a statistical technique called an unrotated one-factor solution showed that the items grouped together with factor loadings ranging between .59 to .81 (see Table 3) thus supporting hypothesis 2. The higher a factor loading, the greater an item contributes to a scale; generally, a loading greater than .3 is considered good (Carmines & Zeller, 1979). The highest loadings were for "My thirst causes me discomfort" and "My thirst bothers me a lot."
Validity was further tested using bivariate correlations with other variables in the conceptual model. The six items of the thirst distress scale were summed so higher numbers indicated more distress. The mean thirst distress score was 17.1 (range 6-30; SD 4.2). A small positive relationship between thirst distress and IWG (r = .17, p < .01) showed some support for hypothesis 3. Only a portion of hypothesis 4 could be tested because all items in the duration and frequency subscales were deleted during item analysis. There was a moderate positive relationship between thirst distress and thirst intensity (r = .31, p < .001).
Preliminary data provide evidence for the reliability and validity of the thirst distress scale. The scale was found to have acceptable content validity, internal consistency reliability, and construct validity. A panel of experts was used to establish content validity and, based upon this review, all items that remained for analysis were content valid. A particular strength of this panel of experts was their diversity of expertise. Their feedback provided information to aid in clarity for the development of each individual item. Item analysis resulted in a six-item thirst distress scale. All items from the duration and frequency subscales were deleted because of low inter-item correlations. If a complete evaluation of the symptom experience of thirst as defined by the Symptom Management Model (UCSF, 1994) is to be conducted, future work will be needed to capture these domains.
Internal consistency reliability of the final six-item scale was satisfactory. The corrected item-total correlations and alpha changes if items were deleted showed sufficient magnitude to warrant retaining the remaining six thirst distress items. Confirmatory and exploratory factor analysis supported the unidimensionality of the scale and indicated significant t-values and acceptable factor loadings for each item.
Further evidence of construct validity was provided by the positive relationships between thirst distress and, respectively, thirst intensity and IWG, which supported the hypothesized relationships. These analyses indicated that the scale did act as theoretically expected. The low correlation between thirst distress and IWG, however, was not expected. We expected thirst distress to be strongly associated with IWG. It is possible, however, that some people do not increase fluid intake just because they are thirsty, that some people do not become particularly thirsty because they drink whatever they want, and/or that other, as yet unknown, factors are important mediators of the thirst-IWG relationship. It is important for nurses to consider all these possibilities when working with the hemodialysis patient.
Preliminary analyses provide support for a summated scale measuring thirst distress with good item-scale correlations. Given the low correlations of the scale with IWG, however, future work is needed to identify factors that are more strongly related to weight gain and, therefore, could be used as a basis for interventions.
Although preliminary data indicate sufficient reliability and validity, further testing with different samples is needed to confirm the findings. The scale could also be tested in patients without ESRD, such as persons with congestive heart failure. Further work could also be done to explore indirect relationships between this variable and fluid management self-care strategies.
There is some indication that hemodialysis patients have more IWG and thirst when they receive a high sodium level during dialysis (Daugirdas, Al-Kudsi, Ing, & Norusis, 1985; Sang, Kovithavongs, Ulan, & Kjellstrand, 1997), although other reports contradict this finding (Cybulsky, Matni, & Hollomby, 1985; Parsons, Yuill, Llapitan, & Harris, 1997). Future work could be done to explore the relationships between sodium profiling during dialysis and the experience of thirst distress to determine whether certain patients are at risk.
The thirst distress scale was initially developed for use in a research study. The scale, however, appears to have potential for use in a clinical setting, particularly in conjunction with measures of other factors that may be related to weight gain. Further testing will help assess clinical usefulness.
Table 1 Correlation Matrix for Items in Thirst Distress Scale Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item 1 1.00 Item 2 .65 1.00 Item 3 .43 .39 1.00 Item 4 .44 .36 .36 1.00 Item 5 .38 .33 .21 .43 1.00 Item 6 .42 .42 .22 .23 .27 1.00 (a) 5/5 5/5 3/5 4/5 3/5 2/5 (b) .46 .43 .32 .36 .32 .31 (a) The proportion or times the item correlations are between .30 and .70 with other items in the scale. (b) The average correlation of the item with all others in the scale. Table 2 Item Statistics for Thirst Distress Scale Corrected Alpha Item-Total if Item M SD Correlation Deleted My thirst causes me discomfort 2.79 1.01 .68 .70 My thirst bothers me a lot. 2.62 .96 .62 .72 I am very uncomfortable when I 3.22 1.04 .45 .76 am thirsty. My mouth feels like cotton when 3.24 1.04 .52 .75 I am thirsty. My saliva is very thick when I 2.99 1.02 .46 .76 am thirsty. When I drink less, my thirst 2.59 .95 .43 .77 gets worse. Table 3 Factor Analysis of the 6-Item Thirst Distress Scale Explora- tory Standardized Confir- Factor Confirmatory matory Item Loadings Coefficients t-values p My thirst causes me .81 .93 18.21 *** discomfort. My thirst bothers me a lot. .76 .91 17.38 *** I am very uncomfortable .66 .62 10.21 *** when I am thirsty. My mouth feels like cotton .66 .55 8.92 *** when I am thirsty. My saliva is very thick .64 .57 9.17 *** when I am thirsty. When I drink less, my .59 .53 8.53 *** thirst gets worse. 48% variance explained *** p < .001
Acknowledgments: This research was supported in part by grants made available to the author by the National Institute of Nursing Research, Center for Enhancing Quality of Life in Chronic Illness (P30 NR05035-01), and the American Nephrology Nurses' Association (ANNA). The views expressed herein are those of the author. No official endorsement by the ANNA is intended or should be inferred.
The author thanks Doris Byers, Barbara Hasbargen, Lydia Carrasquillo, Clifford Blake, and Kim Dixon for their help with data collection; Dr. Victoria Champion for her help with confirmatory factor analysis; Drs. Joan Austin, Victoria Champion, and Phyllis Dexter for their helpful comments and editorial assistance; and the panel of experts who willingly gave their time and expertise when reviewing the proposed items. Finally, appreciation is extended to the participating patients who generously gave their time and energy to this project.
Aroian, K.J., & Norris, A.E. (2001). Confirmatory factor analysis. In B.H. Munro (Ed.), Statistical methods for health care research (pp. 331-354). Philadelphia: Lippincott.
Baldree, K.S., Murphy, S.P., & Powers, M.J. (1982). Stress identification and coping patterns in patients on hemodialysis. Nursing Research, 31, 107-112.
Betts, D.K., & Crotty, G.D. (1988). Response to illness and compliance of long-term hemodialysis patients. ANNA Journal, 15, 96-99.
Boyd, C.J., Frey, M.A., & Aaronson, L.S. (1988). Structural equation models and nursing research: Part I. Nursing Research, 37, 249-252.
Brady, B.A., Tucker, C.M., Alfino, P.A., Tarrant, D.G., & Finlayson, G.C. (1997). An investigation of factors associated with fluid adherence among hemodialysis patients: A self-efficacy theory based approach. Annals of Behavioral Medicine, 19, 339-343.
Carmines, E.G., & Zeller, R.A. (1979). Reliability and validity assessment. Newbury Park, CA: Sage.
Christensen, A.J., Benotsch, E.G., & Smith, T.W. (1997). Determinants of regimen adherence in renal dialysis. In D.S. Gochman (Ed.), Handbook of health behavior research (Volume 2). New York: Plenum.
Cybulsky, A.V.E., Matni, A., & Hollomby, D.J. (1985). Effects of high sodium dialysate during maintenance hemodialysis. Nephron, 41, 57-61.
Daugirdas, J.T., Al-Kudsi, R.R., Ing. T.S., & Norusis, M.J. (1985). A double-blind evaluation of sodium gradient hemodialysis. American Journal of Nephrology, 5, 163-168.
DeVellis, R.F. (1991). Scale development: Theory and applications. Newbury Park, CA: Sage.
Dillman, D.A. (1978). Mail and telephone surveys: The total design method. New York: Wiley.
Dodd, M., Janson, S., Faucett, J., Facione, N., Froelicher, E.S., Humphreys, J., et al. (2001). Advancing the science of symptom management. Journal of Advanced Nursing, 33, 668-676.
Dominic, S.C.R., Ramachandran, S., Somiah, S., Mani, K., & Dominic, S.S. (1996). Quenching the thirst in dialysis patients. Nephron, 73, 597-600.
Everett, K.D., Brantley, P.J., Sletten, C., Jones, G.N., & McKnight, G.T. (1995). The relation of stress and depression to interdialytic weight gain in hemodialysis patients. Behavioral Medicine, 21, 25-30.
Ferketich, S. (1991). Focus on psychometrics: Aspects of item analysis. Research in Nursing & Health, 14, 165-168.
Giovannetti, S., Barsotti, G., Cupisti, A., Morelli, E., Agostini, B., Posella, L., et al. (1994). Dipsogenic factors operating in chronic uremics on maintenance hemodialysis. Nephron, 66, 413-420.
Greenleaf, J.E. (1992). Problem: Thirst, drinking behavior, and involuntary dehydration. Medicine and Science in Sports and Exercise, 24, 645-656.
Gurklis, J.A., & Menke, E.M. (1988). Identification of stressors and use of coping methods in chronic hemodialysis patients. Nursing Research, 37, 236-239, 248.
Long, J.S. (1983). Confirmatory factor analysis: A preface to LISREL. Newbury Park, CA: Sage.
Lynn, M.R. (1986). Determination and quantification of content validity. Nursing Research, 35, 382-385.
Martinez-Vea, A., Garcia, C., Gaya, J., Rivera, F., & Oliver, J.A. (1992). Abnormalities of thirst regulation in patients with chronic renal failure on hemodialysis. American Journal of Nephrology, 12, 73-79.
McIver, J.P., & Carmines, E.G. (1981). Unidimensional scaling. Beverly Hills, CA: Sage.
Nunnally, J.C. (1978). Psychometric theory (2nd ed.). New York: McGraw-Hill.
Oldenburg, B., Macdonald, G.J., & Perkins, R.J. (1988). Factors influencing excessive thirst and fluid intake in dialysis patients. Dialysis & Transplantation, 17, 21-23, 40.
Parsons, D.S., Yuill, E., Llapitan, M., & Harris, D.C. (1997). Sodium modelling and profiled ultrafiltration in conventional hemodialysis. Nephrology, 3, 177-182.
Phillips, P.A., Bretherton, M., Johnston, C.I., & Gray, L. (1991). Reduced osmotic thirst in healthy elderly men. American Journal of Physiology, 261, R166-R171.
Polit, D.F., & Hungler, B.P. (1999). Nursing research: Principles and methods (6th ed.). Philadelphia: Lippincott.
Porth, C.M., & Erickson, M. (1992). Physiology of thirst and drinking: Implication for nursing practice. Heart & Lung, 21, 273-282.
Sang, G.L.S., Kovithavongs, C., Ulan, R., & Kjellstrand, C.M. (1997). Sodium ramping in hemodialysis: A study of beneficial and adverse effects. American Journal of Kidney Diseases, 29, 669-677.
Toto, K.H. (1994). Regulation of plasma osmolality: Thirst and vasopressin. Critical Care Nursing Clinics of North America, 6, 661-674.
Tracy, H.M., Green, C., & McCleary, J. (1987). Noncompliance in hemodialysis as measured with the MBHI. Psychology and Health, 1, 411-423.
United States Renal Data System. (2001). Annual data report. Retrieved September 4, 2001, from www.usrds.org/
University of California, San Francisco School of Nursing Symptom Management Faculty Group. (1994). A model for symptom management. Image: Journal of Nursing Scholarship, 26, 272-276.
Virga, G., Mastrosimone, S., Amici, G., Munaretto, G., Gastaldon, F., & Bonadonna, A. (1998). Symptoms in hemodialysis patients and their relationship with biochemical and demographic parameters. The International Journal of Artificial Organs, 21, 788-793.
Welch, J.L. (2001). Hemodialysis patient beliefs by stage of fluid adherence. Research in Nursing & Health, 24, 105-112.
Welch, J.L., & Austin, J.K. (1999). Factors associated with treatment-related stressors in hemodialysis patients. ANNA Journal, 26, 318-325.
Wirth, J.B., & Folstein, M.F. (1982). Thirst and weight gain during maintenance hemodialysis. Psychosomatics, 23, 1125-1127, 1130-1131, 1134.
Woodtli, A.O. (1990). Thirst: A critical care nursing challenge. Dimensions of Critical Care Nursing, 9, 6-15.
Janet L. Welch, DNS, RN, is a Clinical Nurse Specialist and Assistant Professor at Indiana University School of Nursing, Department of Adult Health, Indianapolis, IN; and the ANNA 2002-2003 Research Chair.
Note: Correspondence concerning this article should be addressed to Janet L. Welch, Department of Adult Health, Indiana University School of Nursing, 1111 Middle Drive, NU 414, Indianapolis, Indiana 46202. E-mail: firstname.lastname@example.org
|Printer friendly Cite/link Email Feedback|
|Author:||Welch, Janet L.|
|Publication:||Nephrology Nursing Journal|
|Date:||Aug 1, 2002|
|Previous Article:||The nephrology nurse's role in improved care of patients with chronic kidney disease.|
|Next Article:||Development of the thirst distress scale. (Commentary and Response).|