How does the onset of incontinence affect satisfaction with life among older women and men? Findings from a nationally representative longitudinal study (German Ageing Survey)

Data were collected as part of the German Ageing Survey (DEAS), which was launched by the German Federal Government in the mid-1990s. It is a large nationwide, longitudinal study of the community-dwelling middle aged and older population in Germany (40 years and over). One of the main aims of this study is to investigate the process of ageing (e.g., health outcomes, or subjective well-being). Due to the cohort-sequential design it is possible to analyze intra-individual change [27].

Since 1996, every 6 years, new nationally representative cross-sectional baseline samples are drawn (n = 4838 in 1996; n = 3670 in 2002; n = 6205 in 2008; n = 6002 in 2014). In addition, longitudinal data for each baseline sample are collected (that is in 2002, 2008, 2011 and 2014). Since the third wave in 2008, the period between waves has been reduced to 3 years. Overall, the response rate was reasonably low in the DEAS study, however it is comparable to other German surveys [28]. The response rates for baseline samples decreased from 50.3% in 1996 to 27.1% in 2014 [27]. The retention rate (valid re-interviews) increased to 41.4% in 2014, based on the 2008 baseline. The sampling methodology and the cohort profile is described in further detail elsewhere [27].

The life satisfaction of individuals was quantified using the well-established Satisfaction with Life Scale (SWLS) [19]. The scale consists of five questions e.g. “In most ways my life is close to my ideal” or “If I could live my life over, I would change almost nothing”, each question is rated on 5-points scale from 1 - “strongly agree” to 5 - “strongly disagree”. The final SWLS score ranges from one to five where high values indicate high life satisfaction [29]. The scale demonstrates favorable psychometric properties [19].

In the DEAS survey, respondents were asked to identify from a list of several illnesses, which illnesses they had been formally diagnosed with by their doctor. Incontinence was included in this list of illnesses. Participants received the list of the illnesses listed with the corresponding letters (A-T) (Appendix). The occurrence of incontinence was determined using responses to this section of the survey (self-reported from participants). The illnesses included in this list were selected based on, among other things, the Charlson Comorbidity Index [30] and consultations with specialists with a background in geriatrics.

With respect to control variables, we included age, gender, marital status (married, living together with spouse; other (divorced; widowed; single)) and labor force participation (working; retired; other: not employed). Furthermore, we controlled for the size of social network (defined as the number of the most important persons with whom individuals are staying in a regular contact with; range from 0 to 9), physical functioning, depression and self-rated health. Physical functioning was measured using the subscale “Physical Functioning” of SF-36 Short Form Health Survey (0–100 range) [31]. High values indicate good physical functioning. Depression was assessed using the Center for Epidemiological Studies Depression Scale (CES-D) [32], ranging from 0 to 45, with high values representing high depressive symptoms with the cut-off at ≥18. Self-rated health is a powerful indicator of clinical outcomes [33], with ratings of overall health ranging from very good (=1) to very poor (=5).

Based on theoretical considerations and empirical studies [9‐14], these variables were selected. Thus, it was adjusted for age, marital status, employment status, number of important people in regular contact, self-rated health, physical functioning and depression in the main model. Stepwise regression models were not used.

In a sensitivity analysis, we extended our model to include level of education: low (0–2), medium (3–4), and high (5–6), according to the International Standard Classification of Education (ISCED-97) [34]. Education level was added to the model as an interaction term with the variable incidence of incontinence. Furthermore, as chronic illnesses are associated with life dissatisfaction [35], the total number of physical illnesses (e.g., diabetes) was added as a surrogate for somatic comorbidity as an independent variable (ranging from 0 to 11). As there is evidence that treatment of prostate cancer is one of the main causes of incontinence in men [5], we excluded individuals with cancer from the regression analysis in a further sensitivity analysis.

Data pertaining to the incidence of incontinence were collected as part of the face-to-face interview. Other variables (e.g. life satisfaction, number of physical diseases) were collected as part of the additional questionnaire completed individually.

Panel regression models were used to examine the association between incontinence and life satisfaction longitudinally. Panel regression models have the advantage of being able to manage time-constant (unobserved) factors such as genetic disposition. These time-constant unobserved factors are treated as random variables in widely used regression models such as random effects (RE) regression models. However, when these unobserved time-constant factors are systematically correlated with the explanatory variables, these estimators will yield inconsistent estimates [36‐38]. In contrast to these models, fixed effects (FE) regression models produce estimates which are consistent when this assumption is violated [36‐38]. For this reason, FE regressions with cluster-robust standard errors [39] were used in the current study. The choice of our model is also motivated by Sargan Hansen tests which correspond to Hausman tests with cluster-robust standard errors. The finding of this test is that the FE estimator is consistent. For example, the Sargan-Hansen statistic was 186.59, p < .001 (main model, total sample), suggesting that there were systematic differences between the coefficients observed using the FE and the RE estimator.

Only changes within individuals over time are used in FE regressions (“Within-estimator”). Therefore, these changes are frequently interpreted in a causal sense (average treatment effect on the treated, ATET). However, the causal inference in our study is restricted because the treatment was not randomly assigned [38, 40].

Given the fact that only changes within individuals over time were used in the FE regressions, observable time-constant factors (e.g., education in older adults or sex) cannot be included as main effects in FE regressions. However, these time-constant factors can be used for group stratification or moderating factors (e.g., sex x incontinence). The potential gender differences were examined by including an interaction term in the main model. We also stratified the analysis by gender. Further details are given elsewhere [38, 40]. The statistical significance was defined as p value of .05 or smaller. Analyses were conducted using Stata 15.0 (StataCorp., College Station, Texas, USA).

Table 1 displays the pooled (i.e., pooled across waves) descriptive characteristics stratified by sex of the observations used for the linear FE regressions with life satisfaction as outcome measure. Altogether 9869 observations (6959 individuals) were included in linear FE regression analysis. In total, 50.6% were male. The mean age in men was 64.9 years (±11.4 years), and the mean age in women was 62.7. In men, average life satisfaction score was 3.8 (±0.7) and in women average life satisfaction score was 3.8 (±0.8). Further details are displayed in Table 1.

It is worth noting that 152 individuals changed their status from ‘absence of incontinence’ to ‘incontinence’ over time. In other words: The number of physician-diagnosed incidental cases of incontinence in the study period was 152 (men: 64; women: 88).

In bivariate FE regression analysis, the onset of incontinence was associated with a decrease in life satisfaction in men (β = − 0.18, p < .05), but not in women (with significant gender differences, p < .05).

Results of multiple FE regression analysis are displayed in Table 2. FE regressions revealed that the onset of incontinence was associated with a decrease in life satisfaction in men (β = − 0.25, p < .01), but not in women. Gender differences (sex x incontinence) were significant (p < .05).

With regard to control variables, a decrease in life satisfaction was associated with onset of depression in the total sample. Individuals who changed their employment status from employed to unemployed also exhibit lower scores of life satisfaction. Additionally, life satisfaction decreased with deterioration of self-rated health. Age was positively associated with life satisfaction. Please see Table 2 for further details (e.g., control variables).

In a sensitivity analysis, it was tested whether the impact of incontinence on life satisfaction differed by educational level. However, none of the interaction terms achieved statistical significance. In a further sensitivity analysis, the main model was extended by adding a sum score of physical illnesses. In a third sensitivity analysis, individuals with cancer were excluded (see Additional file 1). However, in terms of significance and effect size, the impact of incontinence on life satisfaction in men remained almost the same in all models.

Using a nationally representative sample of individuals aged 40 and above, the present longitudinal study examined the onset of incontinence on life satisfaction in the total sample and stratified by sex. Fixed effects regressions revealed that the onset of incontinence was associated with a decline in life satisfaction in men (β = −.25, p < .01), but not in the total sample and in women. Gender differences were significant (p < .05).

Numerous studies have investigated the relation between incontinence and HRQoL [7, 8, 41]. However, it is worth emphasizing that this longitudinal study analyses the impact of incontinence on life satisfaction. Therefore, this section solely focuses on the relation between incontinence and life satisfaction. A few cross-sectional studies have shown that incontinence is associated with decreased satisfaction with life, mainly based on samples including both women and men [9‐11]. Berg et al. [9] investigated life satisfaction (LSI-Z) in the oldest-old (80+). Based on the sample drawn from the Swedish OCTO-Twin study (n = 392) they showed UI is significantly related to decreased life satisfaction. In the mid-80s Herzog et al. [11] showed in a study based on probability sample of 747 women and 541 men aged 60 and older that urinary incontinence is related to low life satisfaction. Zhang and Yu [10] analyzed individuals from a clinical sample (geriatric out-patients; n = 200) and a randomly selected community sample (n = 150). Both samples were from the same area in Beijing. Individuals from the clinical sample with less frequent urinary incontinence reported significantly greater life satisfaction compared to individuals with more frequent symptoms of incontinence. It is worth mentioning that the samples differed significantly, e.g. male gender 63% vs. 36% in the clinical and in the community sample respectively (illiterate 14.5% vs. 46.7%).

While the aforementioned cross-sectional studies found an association between incontinence and life satisfaction, the onset of incontinence was not associated with life satisfaction in the total sample in our study. However, it is difficult to compare these cross-sectional findings with our results, given that our findings are based on nationally representative longitudinal data. These differences may be partially explained by self-selection. Self-selection is well recognized as a key challenge in non-experimental research. Self-selection means that individuals who score low in life satisfaction may have a higher probability of developing incontinence because they had adverse health outcomes prior to the onset of incontinence. FE-regressions, which were used in this study, are able to tackle bias related to self-selection by exploiting intra-individual changes over time.

In our longitudinal study, the onset of incontinence decreased satisfaction with life in men, but not in women, after adjusting for various potential confounders including mental and physical health. These differences may be explained by the fact that the onset of incontinence is a critical life event for men, but not for women. These expected differences in perceptions could be explained by the considerable gender differences in the prevalence of incontinence among older adults. Thus, while men may feel stigmatized when they experience incontinence, women may perceive incontinence rather as an aspect of the natural course of life. In addition, men experiencing incontinence may compare themselves with other men their age who are better off (i.e., do not have incontinence). These negative health comparisons may affect satisfaction with life in men [24]. In contrast, inter-individual comparisons generally did not affect life satisfaction in women.

Results remained virtually the same when individuals with cancer were excluded. Thus, we assume that the impact of incontinence on life satisfaction in men is not driven by consequences of prostate cancer treatment.

Some strengths are worth highlighting. For the present longitudinal study, data were drawn from a large, nationally representative study of community-dwelling individuals in the second half of life. Whether the onset of incontinence was associated with a change in life satisfaction over the course of 6 years was examined. Life satisfaction was measured using a well-established scale. Unobserved heterogeneity, a main challenge in large survey studies and life satisfaction research, was addressed by using FE regressions in the current study. Unlike previous studies based on self-reported incontinence, physician-diagnosed presence of incontinence was reported by the individuals in this study.

On the other hand, some limitations are worth noting. A small sample selection bias have been observed in the DEAS study. Though quite small, panel attrition was observed in this study. Nevertheless, intense efforts have been made to decrease attrition, and the retention rate has increased. Please see Klaus et al. [27] for further details. The distribution of sociodemographic factors in the sample is very close to that of the German population [42]. Due to the sensitive question concerning incontinence, bias may have occurred. The physician-diagnosed presence of incontinence was quantified in the DEAS study without distinguishing between UI and FI. However, only approximately 6.0% of women and 1.9% of men reported both UI and FI [4]. Furthermore, the prevalence of FI, without symptoms for both FI and UI, is estimated to be about 8%, for both community-dwelling women and men [4]. Nevertheless, further research is required that considers the severity of urinary and fecal incontinence. Moreover, other time-varying factors (e.g., cognitive decline, walking speed) may affect our relationship of interest. Future research is required to examine these factors.