Predictors of Attrition in a Large Clinic-Based Weight-Loss Program

Transcription

1 Predictors of Attrition in a Large Clinic-Based Weight-Loss Program Jeffery J. Honas,* James L. Early,* Doren D. Frederickson,* and Megan S. O Brien* Received for review August 14, Accepted in final form May 14, *University of Kansas School of Medicine-Wichita and Via Christi Regional Medical Center, Wichita, Kansas. Address correspondence to James L. Early, Department of Preventive Medicine and Public Health, University of Kansas School of Medicine-Wichita, 1010 N. Kansas, Wichita, KS jearly@kumc.edu Copyright 2003 NAASO Abstract HONAS, JEFFERY J., JAMES L. EARLY, DOREN D. FREDERICKSON, AND MEGAN S. O BRIEN. Predictors of attrition in a large clinic-based weight-loss program. Obes Res. 2003;11: Objective: Identifying client factors that predict dropout is critical for the development of effective weight-loss programs. Although demographic predictors are studied, there are few consistent findings. The purpose of this study was to identify predictors of dropout in a large clinic-based weightloss program using readily attainable demographic variables. Research Methods and Procedures: All 866 weight-loss patients in a clinic-based weight-loss program enrolled during 1998 to 1999 were followed. Attrition and retention rates were measured at 8 and 16 weeks. Six variables (sex, race, marital status, age, BMI, and treatment protocol) were evaluated using bivariate and multivariable statistics for relative association with dropout. Results: The overall attrition rate for the 16-week program was 31%. The retention rate was 69%. Significant risk for dropout, measured as bivariate relative risk (95% confidence interval), was found among patients who were: females, 1.32 (1.01 to 1.73); divorced, 1.54 (1.13 to 2.09); African Americans, 1.68 (1.26 to 2.23); age 40, 1.66 (1.27 to 2.18); and ages 40 to 50, 1.33 (1.01 to 1.76). There were no significant differences in retention rates by BMI group or program protocol. After logistic regression analysis to control for all variables, young age 50 years had the only significant association with dropout [odds ratio 1.39 (1.02 to 1.90)]. Discussion: Multivariable modeling was helpful for prioritizing risk factors for program dropout. These findings have important implications for improving weight-loss program effectiveness and reducing attrition. By knowing the groups at risk for dropout, we can improve or target program treatments to these populations. Key words: dropout, weight loss, cohort studies, predictor, causality Introduction It is well established that weight-loss programs of different types can help individuals lose at least 10% of their weight. However, for treatment to be effective, the patient must first complete the program. In an important study of hospital-based clinics in the U.S., Wadden et al. (1) found attrition rates as high as 46%. Unfortunately, program attrition results in failure of many individuals to achieve their weight-loss goals, which may limit the utility of the program. Many variables have been investigated as predictors of attrition from weight-loss programs (2 17). Davis and Addis (2) reviewed 13 behavioral weight-loss program attrition studies in which over one-third (33%) of the variables examined were demographic (age, marital status, occupation, race, education, etc.). Psychological and behavioral variables were the second and third most widely studied (27% and 20%, respectively). Some of the variables that have been found to be associated with lower attrition from weight-loss programs include having less depression, being a nonsmoker, an exerciser, and older, and having greater number of past diets, fewer past physical and emotional problems, greater expectation of non-weight-loss-related stress in the next 6 months, and less expectation of insurance coverage (4,12). Studies have varied by sample size, type of program, length of treatment, and definition of attrition. Not surprisingly, no common attrition tendency has been found (4,2,7). Studies have often differed by exclusionary criteria, such as diabetes (8), heart disease, or the taking of antidepressant medication (5). These issues may limit the generalizability 888 OBESITY RESEARCH Vol. 11 No. 7 July 2003

2 of study findings and their usefulness in the development of attrition interventions for real world clinic-based weightloss clinics. Clinic-based weight-loss programs often do not have the resources to collect data beyond demographics, vital statistics, and self-reported energy intake/expenditure. Thus, although research has shown that psychological variables such as depression are predictive of attrition (4), properly administered and evaluated psychological assessments may be a task impossible for clinic-based programs. It makes clear the pressing need for practical demographic variables that can be easily identified on entry. Clearly, it is important for attrition studies to be representative of the patients who normally seek out or are referred to weight management treatments. The purpose of this study was to evaluate the overall dropout rate and demographic dropout patterns of overweight and obese patients in a clinic setting with a diverse population broadly representative of the community. This study examined six variables (sex, race, marital status, age, BMI, and treatment protocol) as predictors of attrition and retention in a large clinic-based weight-loss program. Research Methods and Procedures Participants The subjects were 866 adult weight-loss patients who entered the weight management program at a Midwest clinic-based medical center over a 2-year period (1998 to 1999). Table 1 displays demographics of the population. The patients entering the program were predominately female (73%) and white (91%). Sixty-two percent were married, 22% were single, and 9% were divorced. BMI had a bimodal distribution; 40% of patients had a BMI 35, 18% had a BMI 35 to 40, and 42% had a BMI 40. The mean age of the patients was 47. Treatment Protocol Before entering the weight management program, all patients were given a complete physical exam. Patients were assigned to one of four protocols based on BMI (kilograms per meters squared) and comorbidities (Table 2). All patients were prescribed a minimum energy intake between 500 and 1200 kcal/d using Health Management Resources meal replacements (shakes and packaged meals) according to the treatment protocol. The program consisted of groups ranging from 12 to 20 participants. The topics discussed in the groups included knowledge development in physical activity and nutrition and personal environmental problem solving. The majority of the class time was spent on weekly planning and skill-building exercises. The behavioral educators in the program encouraged patients to achieve levels of physical activity averaging over 2000 kcal/week in the weight-loss program. The liquid supplements were very low Table 1. Demographics of study population (N 866) Variable n % of total Sex Female Male Marital status Married Single Divorced 77 9 Missing 60 7 Race White African American 54 6 Other 24 3 Age to to BMI BMI BMI35to BMI Treatment protocol Protocol Protocol Protocol Protocol Attrition Dropped out Completed program fat with a balance of quality protein and carbohydrate; they ranged from 104 to 160 kcal per supplement. Meals were retort packaged with low fat content and a balance of protein and carbohydrate and averaged 200 kcal per prepackaged meal (18,19). Physician visit frequency and minimum daily energy intake differentiated the four protocols. The weight-loss program consisted of 16 weekly behavioral meetings, which were 1.5 hours long. Participants had the option of choosing between an evening and a daytime class. The group health educators followed the Health Management Resources program guide during the weekly behavioral meetings. All patients attended the same class structure during the 16- week weight-loss phase. Patients were assigned to a protocol based on accepted clinical guidelines (20). Patients assigned to Protocol I had OBESITY RESEARCH Vol. 11 No. 7 July

3 Table 2. Summary of treatment protocol BMI (kg/m 2 ) Level of medical supervision Minimum kilocalories Level of medical risk Screening procedure Protocol 1 40 Weekly physician monitoring 8 weeks, then biweekly Protocol 2 35 and 40 Biweekly physician monitoring 8 weeks, then every 4 weeks Protocol 3 35 Nurse visits every 4 weeks Protocol 4 35 No medical supervision by program medical staff* 520 kcal Patients with diabetes on insulin or with poorly controlled hypertension, and other patients judged to be at high risk during screening 800 kcal Patients with diabetes on oral agents 800 kcal Patients without diabetes or other unstable medical problems 1200 kcal including fruits, vegetables, and prepackaged meals Patients without diabetes or other unstable medical problems Complete history and physical by program nurse and physician Complete history and physical by program nurse and physician Complete history and physical by program nurse and physician Nurse intake with questionnaire and basic body composition with vital sign measurements * Letter to primary physician asking them to provide medical oversight. Additional calories in liquid supplement and/or prepackaged meals added at the discretion of the medical/behavioral team. Option for patients with BMI 35 and low medical risks. Patients without medical risks but BMI 35 could choose between Protocols 3 and OBESITY RESEARCH Vol. 11 No. 7 July 2003

4 abmi 40 and were required to see the physician weekly. They were prescribed a minimum of 500 kcal/d. Patients assigned to Protocol II had a BMI 35 to 40 and were required to have physician contact on a biweekly basis. They were prescribed a minimum of 800 kcal/d. Patients with a BMI 35 who did not have medical risks such as high blood pressure or diabetes had a choice of receiving medical supervision (Protocol III) or no medical supervision (Protocol IV) during the weight-loss phase. However, patients with a BMI 35 with medical risks needing medical supervision were assigned to Protocol III. They were prescribed a minimum of 800 kcal/d and were not normally seen by the physician but were supervised by the nurse health educator weekly. Patients assigned to Protocol IV had a BMI of 35. They were prescribed a minimum of 1200 kcal/d, which included fruits and vegetables. Those in Protocol IV group were not medically supervised. Analyses All patients attending the first class were included for analyses. Dropout rates were calculated at 8 and 16 weeks. The dropout rate at week 8 was defined as leaving the weight-loss program before week 9. The week 16 dropout rate was defined as leaving the weight-loss program after week 8 and before 16. The total dropout rate was defined as leaving the program any time before week 16. We used early and late dropout intervals from prior research as a guide (12). Patients who missed more than two classes at any time during the 16 weeks were also considered dropouts and were released from the program. Six variables (sex, race, BMI, marital status, age, and treatment protocol) were evaluated in characterizing the dropout rates. SPSS 9.0 (21) was used for descriptive statistics. Significant risk of dropout was measured as bivariate relative risk [95% confidence interval (CI)] using Epi Info 6.04 (22). Logistic regression analysis was done using SAS 10.0 (23). Results The overall dropout rate for the 16-week program was 31%. Bivariate risk for dropout measured by risk ratio (95% CI) was found among patients who were: female, 1.32 (1.01 to 1.73); divorced, 1.54 (1.13 to 2.09); African American, 1.68 (1.26 to 2.23); age 40, 1.66 (1.27 to 2.18); and age 40 to 50, 1.33 (1.01 to 1.76) (see Table 3 and Figures 1 to 4). There were no significant differences in dropout among BMI groups and treatment protocol groups (See Table 3). Table 4 displays the odds of dropout after logistic regression analysis. After controlling for other variables, those patients with age 50 years were the only group with a significant risk for dropout [odds ratio 1.39 (95% CI 1.02 to 1.90)]. Discussion Considering the increased national prevalence of obesity and its health consequences, it has become a matter of Table 3. Bivariate risk of drop out Risk factor Relative risk (95% CI) p Sex Female 1.0 ref Male 1.32 (1.01 to 1.73) 0.04 Marital status Married 1.0 ref Single 1.21 (0.94 to 1.56) NS* Divorced 1.54 (1.13 to 2.09) 0.01 Race White 1.0 ref African American 1.68 (1.26 to 2.23) Other 1.26 (0.74 to 2.213) NS Age (1.27 to 2.18) to (1.01 to 1.76) to ref (0.86 to 1.85) NS BMI (kg/m 2 ) BMI ref BMI35to (0.83 to 1.43) NS BMI (0.83 to 1.29) NS Treatment protocol Protocol (0.82 to 1.38) NS Protocol (0.86 to 1.57) NS Protocol (0.94 to 1.76) NS Protocol ref * NS, not statistically significant. ref, reference. urgency to focus on increased effectiveness of treatment programs. To be effective, weight control programs must either prevent obesity or treat established overweight and obesity and then maintain the improved weight status. Without completion of an initial weight-loss and basic education program, it is difficult for patients to learn and practice a comprehensive set of behavioral and knowledge-based skills. There are behavioral and psychological reasons for success and failure in weight management programs. However, these variables exist across all demographic subgroups. Our focus in the study of adult weight-loss patients was to characterize the demographic predictors of program dropout as a basis for program improvement and further research. Previous studies have disagreed about the relationship of obesity to program attrition (10,11). In our study, comparison by initial BMI was accomplished by comparing those patients with Level III obesity to those with Level II OBESITY RESEARCH Vol. 11 No. 7 July

5 Figure 1: Retention rates across a 16-week weight-loss program, by gender. *, p 0.05; **, p 0.01; ***, p Figure 3: Retention rates across a 16-week weight-loss program, by race. *, p 0.05; **, p 0.01; ***, p and Level I obesity or overweight (20). Dropout rates were not different among BMI groups in this comparison. Overall dropout rates in this study were typical of those seen in other behaviorally based weight management programs (1,2). Several investigators have hypothesized that with a more intense treatment program (for example, a low-calorie program), those with a higher BMI would have lower dropout rates. In contrast, those with a lower BMI may have less attrition from a less intensive program (a 1200-calorie program) (9). However, our finding of no relationship between BMI and attrition suggests that this was not the case. In our study, all patients received the same behavioral and knowledge-based program. The current findings suggest that if patients are given the same behavior-based treatment, BMI does not influence attrition. Interestingly, neither race nor marital status was associated with dropout rates in the multivariable analysis. This intriguing finding was exactly the opposite of what the research team had hypothesized. Although the bivariate analysis showed several variables (African-American race, divorce, and female sex) to display a trend toward or have significant association with attrition, the logistic regression multivariable analysis showed that these were not significantly associated with attrition after all variables entered the model. We concluded that, in this clinic population, these covariates operated as confounders and not as primary risk factors associated with attrition. This finding may result from special attention paid to social, racial, and gender cultural sensitivity issues that inspired patients from those groups to remain in the weight-loss program or may result from other reasons unknown to the investigators. There are a limited number of studies that have examined predictors of obesity program attrition. Strengths of the new study presented here were the large sample size and the use Figure 2: Retention rates across a 16-week weight-loss program, by marital status. *, p 0.05; **, p 0.01; ***, p Figure 4: Retention rates across a 16-week weight-loss program, by age. *, p 0.05; **, p 0.01; ***, p OBESITY RESEARCH Vol. 11 No. 7 July 2003

6 Table 4. Logistic regression model for dropout Risk factor Odds ratio (95% CI) p Sex Male 1.0 ref Female 1.40 (0.93 to 2.11) NS* Marital status Married with spouse 1.0 ref Single/divorced/widowed 1.30 (0.93 to 1.81) NS Race White 1.0 ref Nonwhite 1.52 (0.90 to 2.56) NS Age ref (1.02 to 1.90) 0.04 BMI BMI ref BMI35to (0.88 to 2.09) NS BMI (0.81 to 1.63) NS * NS, not significant. ref, reference. Model strength: Hosmer-Lemeshow c statistic Model strength: Wald Test p of multivariable statistics to control cofounders. To fully appreciate the present findings, however, it is important to address several limitations. It is possible that the low caloric intake and the use and role of prepacked meals may offer several incentives in terms of ease, structure, and a decrease in decisional anxiety. Although an increase in structure may, to some degree, limit the generalizability, this limitation is somewhat mitigated by the fact that the use of prepacked meals is becoming an accepted tool. Another possible limitation is the seemingly short-term nature of follow-up. However, this was not designed to be an assessment of long-term adherence but rather an assessment of success, measured by retention, of a real world clinic-based program with a standard 16-week protocol. These results are important in light of the fact that adherence to a core program at 8 and 16 weeks is critical for setting the stage for long-term adherence. In this study, we measured attrition at 8 and 16 weeks, anticipating that risk factors for attrition among a clinic population might change over time as patients face shifting medical and social demands of the weight-loss process. As the patterns displayed in Figures 1 to 4 show, we were fascinated to discover that no crossover in stratum-specific attrition rates occurred. Although we hypothesized that there may be different reasons for dropout early or late in a weight-loss program, our study showed that dropout rates for the first 8 program weeks mirrored those of the final 8 weeks across all BMI-based protocols, regardless of the intensity of medical supervision or calorie level. The stable linearity of attrition risk among strata, evidenced by the absence of crossover, suggested the presence of stable risk effects. Of course, it is possible that risk factors may differ at other times during the program. Risk factors for attrition might well be different when patients are followed out to 1 or even 2 years. In fact, 2 years is the actual duration of the clinic weight-loss program studied by this research group. However, we chose 4 months (16 weeks) because these were felt to be critical months when program adherence is established, because 4 months is the norm for many weight management programs, and because 4 months is the precedent for long-term success. Age appeared to be the most significant demographic determinant of drop rate in our study; this finding corresponded with previous research (4). In our cohort, 76% of 51- to 60-year-old participants completed the weight program, whereas only 60% of participants under 40 years completed the program. There are a number of theoretical reasons that older participants may have better sustained participation in weight-loss programs. Declining health may motivate the older cohort, and their participation may be enhanced by fewer family obligations. However, the desire for social success and enhanced attractiveness might have been expected to motivate the younger cohort more strongly. In light of these findings and in light of current trends in obesity in the young, it appears critical that we enhance our approach to prevention and treatment especially for those under 40 years (24). The results obtained should now encourage us to begin to probe for ways to enhance program effectiveness for younger patients while continuing to maintain and improve completion rates in older patients. Acknowledgments There was no outside funding/support for this study. References 1. Wadden TA, Foster GD, Letizia KA, Stunkard AJ. A multicenter evaluation of a proprietary weight reduction program for the treatment of marked obesity. Arch Int Med. 1992;152: Davis MJ, Addis ME. Predictors of attrition from behavioral medicine treatment. Ann Behav Med. 1999;21: Wadden TA, Stunkard AJ. Controlled trial of very low calorie diets, behavior therapy, and their combination in the treatment of obesity. J Consult Clin Psychol. 1986;54: Clark MM, Niaura R, King TK, Pera V. Depression, smoking, activity level, and health status: pretreatment predictors of attrition in obesity treatment. Addictive Behav. 1996;21: Ho KSI, Nichaman MZ, Taylor WC, Lee ES, Foreyt JP. Binge eating disorder, retention, and dropout in an adult obesity program. Int J Eating Disord. 1995;18: OBESITY RESEARCH Vol. 11 No. 7 July

7 6. Mitchell C, Stuart RB. Effect of self-efficacy on dropout from obesity treatment. J Consult Clin Psychol. 1984;6: Pratt CA. Factors related to the completion of a weight reduction program. J Am Dietetic Assoc. 1990;90: Bennett GA, Jones SE. Dropping out of treatment for obesity. J Psychosom Res. 1986;30: Pekarik G, Blogett C, Evans ER, Wierzbicki M. Variables related to continuance in a behavioral weight loss program. Addictive Behav. 1984;9: Clark MM, Guise BJ, Niaura RS. Obesity level and attrition: support for patient-treatment matching in obesity treatment. Obes Res. 1995;1: Ek A, Anderson I, Barkeling B, Rossner S. Obesity treatment and attrition: no relationship to obesity level. Obes Res. 1996;3: Yass-Reed EM, Barry NJ, Dacey CM. Examination of pretreatment predictors of attrition in VLCD and behavior therapy weight-loss program. Addictive Behav. 1993;18: Fower JL, Follick MJ, Abrams DB, Rickard-Figueroa K. Participant characteristics as predictors of attrition in worksite weight loss. Addictive Behav. 1985;10: Ashley JM, St Joer ST, Schrage JP, et al. Weight control in the physician s office. Arch Intern Med. 2001;161: Mavis BE, Stoffelmayr BE. Multidimensional evaluation of monetary incentive strategies for weight control. 1994;44: Feuerstein M, Papciak A, Shapiro S, Tannenbaum S. The weight loss profile: a biopsychosocial approach to weight loss. Int J Psych Med. 1989;2: Fontaine KR, Cheskin LJ. Self-efficacy, attendance, and weight loss in obesity treatment. Addictive Behav. 1997;4: Anderson JW, Hamilton CC, and Brinkman-Kaplan V. Benefits and risks of an intensive very-low-calorie diet program for severe obesity. Am J Gastroenterol. 1992;1: Health Management Resources, Inc. Health Management Resources Medical Guidelines. Boston, MA: Health Management Resources, Inc.; Expert Panel on the Identification, Evaluation, and Treatment of Overweight in Adults. Clinical Guidelines on the Identification, Evaluation, and Treatment of Overweight and Obesity in Adults: executive summary. Expert Panel on the Identification, Evaluation, and Treatment of Overweight in Adults. Am J Clin Nutr 1998;68: SPSS Inc. SPSS 9.0. Chicago, IL: SPSS Inc., Centers for Disease Control and Prevention. Epi Info 6.04b. Atlanta, GA: Centers for Disease Control and Prevention; SAS Institute Inc. SAS 8.1. Cary, NC: SAS Institute Inc.; Mokdad AH, Bowman BA, Ford ES, Vinicor F, Marks JS, Koplan JP. The continuing epidemics of obesity and diabetes in the United States. JAMA. 2001;286: OBESITY RESEARCH Vol. 11 No. 7 July 2003