The respective roles of negative and positive views of aging in predicting rehabilitation outcomes

Abstract

Objectives

Subjective views of aging (VoA) predict health outcomes, yet most studies rely on specific measures of VoA in relatively healthy samples. Referring to multidimensional (generalized and personal) and multidirectional (negative and positive) VoA constructs, we examined whether these VoA constructs differentially predict rehabilitation outcomes among older adult patients coping with hip fracture and stroke.

Methods

158 geriatric rehabilitation patients (mean age = 77.15, SD = 6.71) reported VoA at admission while referring to negative and positive generalized (Attitudes to Aging Questionnaire, AAQ) and personal views (awareness of age-related changes, AARC). Patients’ functional independence was evaluated by healthcare personnel at admission and discharge, and rehabilitation efficiency (RE) was computed to reflect functional improvement relative to potential. Path analyses tested the unique contributions of negative and positive VoA (for both generalized and personal views) while adjusting for other predictors of RE.

Results

For both generalized and personal VoA, negative views (i.e., AAQ-Losses and AARC-Losses) predicted lower RE, whereas positive views (i.e., AAQ-Growth and AARC-Gains) did not show significant effects on RE. Negative views had a significantly stronger effect on RE than positive views, but only for the AAQ subscales.

Discussion

In the rehabilitation context, negative VoA, referring to age-related losses, appears more consequential for functional recovery than positive VoA (referring to age-related growth and gains). Findings support prioritizing interventions that identify and modify both generalized and personal attitudes, especially those focused on age-related losses, during rehabilitation to enhance functional outcomes.

In recent decades, the concept of subjective views of aging (VoA) gained increased popularity in the gerontological literature (Diehl et al., 2014). VoA is an overarching term that refers to “individuals’ perceptions, attitudes, or expectations of their own age and aging process, as well as of old age and older adults in general” (Shrira et al., 2022, p. 11). VoAs are also multidirectional (Diehl et al., 2014; Kornadt et al., 2020), meaning that they encompass both negative and positive views, focusing on age-related losses and gains, respectively (Miche et al., 2014). Moreover, recent conceptualizations divide VoA into generalized views, which capture attitudes toward old age and older adults in general (e.g., age-related stereotypes; Rothermund & de Paula Couto, 2024), and personal views, which represent perceptions of one’s own aging (Diehl & Wahl, 2024).

The concept of VoA gained popularity due to the accumulation of substantial evidence linking VoA to multiple health outcomes, including longevity (Westerhof et al., 2023; Wurm et al., 2017). Additionally, VoA has recently been conceptualized as pivotal to successful aging (Sabatini et al., 2025). However, most works used unidirectional measures (e.g., subjective age), so the respective effects of negative and positive VoA were seldom compared (Westerhof et al., 2023). Indeed, a recent meta-analysis showed that only 13% of longitudinal studies examined both negative and positive VoA when predicting health outcomes (Westerhof et al., 2023). Also, despite the vast evidence showing that VoA predicts subsequent health outcomes, studies mainly focused on either generalized or personal views and rarely examined the potential effect of both dimensions (Diehl et al., 2014; Wurm et al., 2017).

To overcome the aforementioned limitations, this study aimed to capture the multidirectionality and multidimensionality of VoA when assessing their predictive role for health outcomes, with a focus on rehabilitation outcomes among older adult patients with orthopedic and neurologic conditions. More specifically, this study referred to two subtypes of VoA—attitudes toward aging and awareness of age-related changes—accounting for their valence, assessing both negative and positive views, and capturing both generalized and personal views. Below, we elaborate on each of these VoA constructs and their potential role in the rehabilitation setting.

The first VoA construct referred to in this study, attitudes toward aging, encompasses perceptions of old age and the aging process (Diehl et al., 2014). Current operationalizations, especially the Attitudes to Ageing Questionnaire (AAQ; Laidlaw et al., 2007), enable a multidirectional assessment, including perceptions focused on age-related losses (e.g., old age is a depressing time of life) and positive views, such as physical change, reflecting old age as a period where one feels healthier than expected (e.g., it is possible to preserve energy and health in old age) and age-related growth (e.g., there are pleasant things in growing old).

The second construct, awareness of age-related change (AARC), refers to the awareness that various personal experiences, such as behavior, performance, and ways of experiencing life, have changed with age (Diehl & Wahl, 2010). It is also multidirectional in the sense that it includes awareness of age-related losses (e.g., realizing one has to limit one’s activities as one gets older) and gains (e.g., realizing one appreciates close relationships much more as one gets older) (AARC-Losses and AARC-Gains, respectively) (Diehl & Wahl, 2010; Miche et al., 2014).

AAQ and AARC represent two dimensions of VoA; the former is characterized by generalized views that reflect stereotypes,¹ while the latter strictly concerns personal views (Diehl et al., 2014; Shrira et al., 2022). Following the stereotype embodiment theory (Levy, 2009), generalized VoA forms earlier in life and therefore affect personal VoA in a process of internalization. However, it is also possible that personal views “project out” and modify generalized views in a process of externalization (Rothermund & Brandtstädter, 2003). Because how personal and generalized VoA links with health have not been extensively studied, it is important to address this issue.

Although many studies have focused on specific VoA constructs, some evidence suggests that different constructs may exhibit unique effects, and some may serve as more potent predictors of health outcomes than others. As noted, very few studies included both positive and negative VoA and generalized and personal VoA when predicting subsequent health outcomes. Regarding the valence of VoA, previous findings show that AAQ subscales referring to losses and physical change, but not to growth, predicted subsequent self-rated physical health among older adults with physical multimorbidity (Bodner et al., 2023). However, in another study, only the physical change subscale predicted lower mortality (McLachlan et al., 2020). Negative VoA showed a more consistent pattern of more substantial effects when referring to AARC. One meta-analytic study demonstrated associations between AARC losses and both physical and emotional outcomes, while AARC gains were mainly associated with emotional outcomes (Sabatini et al., 2020). Moreover, while both AARC-Losses and AARC-Gains predicted mortality, only AARC-Losses remained a significant predictor of higher mortality after controlling for various health and psychosocial variables (Kaspar et al., 2021).

Regarding the differences between generalized and personal VoA, a recent study found that both predicted mortality in distinct ways. Personal VoA was positively linked with mortality (the more positive VoA was, the longer people lived). Generalized VoA, however, showed the opposite association; the less positive the VoA was, the longer people lived (de Paula Couto et al., 2025). These authors suggested that negative generalized stereotypes can reflect a dissociative position, in which one distances oneself from older adults to protect oneself.

As mentioned above, the current study aimed to assess the role of various VoA among older adult patients with orthopedic and neurologic conditions. Notably, most studies focus on VoA of healthy individuals as predictors of future health outcomes. However, VoA may also influence coping and recovery from medical conditions; yet, few studies have assessed VoA in clinical samples of older adults (Wurm et al., 2022), and even fewer studies have examined older adults during hospitalization and rehabilitation (Kalir et al., 2023; Zisberg et al., 2024).

This latter point is not trivial, as the VoA of older adults who cope with medical conditions (e.g., patients in a hospital setting) may have little predictive power, as measurements are conducted on people whose physical and psychological status is suboptimal, suffering, and in significant decline. Thus, such recorded VoA are most likely influenced by deteriorated health and may be less potent predictors of future recovery or rehabilitation. On the other hand, VoA may be robust enough to predict future recovery even in such cases. In the very few studies that did assess VoA in hospital settings, a younger subjective age was associated with better post-hospitalization health outcomes (Zisberg et al., 2024) and improved rehabilitation outcomes among older patients with orthopedic and neurologic conditions (Kalir et al., 2023).

To extend previous studies, we aimed herein to account for the multidirectionality and multidimensionality of VoA. Building on previous findings, we hypothesized that the effect of negative VoA on rehabilitation outcomes would be stronger than that of positive VoA. There was insufficient evidence to guide a specific hypothesis regarding the relative effects of generalized (AAQ) and personal (AARC) VoA.

In summary, this study has three goals: first, to assess whether VoA during hospital admission predicts rehabilitation outcomes 6–8 weeks later; second, to examine whether the previously reported advantage of negative VoA in predicting health outcomes would extend to a rehabilitative context; and third, to explore whether there are differences between personal and generalized VoA in predicting rehabilitation outcomes (no hypothesis could be made due to a dearth of previous studies), with VoA recorded from older adults right after their significant decline (stroke or hip fractures).

Method

Participants and procedure

The sample comprised 158 patients undergoing rehabilitation after a hip fracture or stroke. The mean age was 77.15 years (SD = 6.71), with a range of 64 to 98 years. Among them, 39.6% were male, 50.9% married, 27.7% widowed, 13.8% divorced, and 8.6% single. Participants had an average of 11.70 years of education (SD = 3.28). Of the total, 124 patients had experienced a hip fracture, while 34 had suffered a stroke. The average hospitalization duration in rehabilitation was 36.96 days (SD = 19.94). According to the criterion of 20 participants per variable (Buhi et al., 2007), at least 160 participants were required to test the model with up to eight variables.

When comparing patients with stroke to those with fractures, we found few differences in background characteristics and no differences in the main study variables. Relative to patients with stroke, patients with fractures were older, had a higher proportion of women, and, as expected, had a higher score on the Mini-Mental State Exam (MMSE). The two groups did not differ on any other variables (see Supplementary Methods; Table S1). Although stroke and hip fractures are different health issues with likely varying risk factors, underlying causes, and resulting complications (Kramer et al., 1997), patients with either condition face significant functional decline. Thus, we grouped these two conditions in all analyses while adjusting for condition type (stroke vs fracture).

The study was conducted at two geriatric rehabilitation hospitals in Israel. All patients had been transferred from acute care wards to these facilities. Inclusion criteria included being at least 60 years old, hospitalized for stroke or hip fracture, and cognitively intact (MMSE score above 23; Folstein et al., 1975). The research assistants first received the identification details for each admitted patient and verified that each patient met the previously specified research inclusion criteria. Subsequently, the research assistants met with each patient to explain the study and its requirements and asked whether they were interested in participating after discussing the study’s goals and requirements. Patients who agreed signed a written informed consent. Ethical approval was secured from the Helsinki committees at both sites. Recruitment occurred from May 2022 to November 2024.

Measures

Attitudes toward aging were assessed using the short version of the Attitudes to Ageing Questionnaire (AAQ; Laidlaw et al., 2018). The scale has 12 items ranging from 1 (strongly disagree) to 5 (strongly agree). It has three subscales, one assessing psychosocial loss (“I see old age mainly as a time of loss”), the second assessing physical change (“I have more energy now than I expected for my age”), and the third assessing psychological growth (“It is a privilege to grow old”). Ratings in each subscale were averaged across items. Laidlaw et al. (2018) reported the following Cronbach’s α for the AAQ: 0.72 for both the subscales of psychosocial loss and physical change and 0.62 for the subscale of psychological growth. In the current study, Cronbach’s α for the Hebrew version was 0.68, 0.59, and 0.66 for the psychosocial loss (AAQ-Loss), physical change (AAQ-Physical), and psychological growth (AAQ-Growth) subscales, respectively.

Awareness of age-related changes (AARC) was assessed using the 10-item short-form questionnaire developed by Kaspar et al. (2019). Each item was rated on a scale ranging from 0 (not at all) to 5 (very much). Ratings were averaged across items referring to losses (e.g., “with my increasing age I realize that I find it harder to motivate myself”) and gains (e.g., “with my increasing age I realize that I have a better sense of what is important for me”) (five items for each subscale), with higher scores reflecting higher loss- and gain-related AARC, respectively. Previous studies found good internal reliability with Cronbach’s α of 0.89 for AARC-Losses and 0.88 for AARC-Gains (Kaspar et al., 2019). In our study, Cronbach’s α for the Hebrew version was lower but still acceptable: 0.77 for AARC-Losses and 0.62 for AARC-Gains.

The rehabilitation outcome was based on scores from the Functional Independence Measure (FIM) (Keith et al., 1987). The FIM is an accepted and validated tool for measuring patients’ functional status in a rehabilitation context (Pollak et al., 1996). The FIM was developed in the United States during the 1980s and has been translated into many languages, including Hebrew (Kalir et al., 2023). This measure was used to quantitatively assess patients’ independence in performing their daily activities. The instrument comprises 18 items, rated by a multidisciplinary team on a scale of independence. The domains rated in FIM include sphincter control, mobility, locomotion, and social cognition, with each domain scored on a scale of 1 to 7 (1 = total dependence, 7 = total independence). The FIM was rated at admission and later at discharge. Each rating was summed to yield a score ranging from 18 to 126. In the current study, Cronbach’s α was 0.86 and 0.90 for FIM at admission and discharge, respectively.

The rehabilitation effectiveness (RE) score (also known as the Montebello Rehabilitation Factor score, Koh et al., 2013) was calculated as follows: (Discharge FIM—Admission FIM)/(Maximum possible FIM score—Admission FIM). Therefore, the RE score reflected the ratio of actual improvement to potential maximum improvement. In contrast to the standard FIM delta score (Discharge FIM—Admission FIM), which indicates only the rehabilitation achieved, the RE score evaluates the extent to which rehabilitation progress has helped the patient toward complete independence (Koh et al., 2013).

Other variables included age, gender, education (in years), family status (re-coded as living with or without a partner), and subjective health (asking patients to rate their health over the last six months on a scale ranging from 1 = not good at all to 5 = very good). The Mini-Mental State Exam (Folstein et al., 1975) was administered at admission.

Data analysis

As mentioned earlier, the first hypothesis was that VoA would predict rehabilitation outcomes. Second, we hypothesized that negative VoA would be a stronger predictor of rehabilitation outcomes than positive VoA. Because of the gap in existing literature, we could not formulate a hypothesis about differences in the effects of generalized vs personal VoA.

To test the study hypotheses, we conducted path analyses with AMOS 29.0. These analyses evaluated the effect of VoA during the first week of rehabilitation on RE. In one model, we included the AAQ subscales; in the second, the AARC subscales. Models controlled for age, gender, education level, family status, subjective health, the primary reason for rehabilitation (stroke vs fracture), and the Mini-Mental State Exam score. These factors were controlled because they could impact rehabilitation outcomes (Koh et al., 2013).

Following Hu and Bentler’s (1999) guidelines for small sample sizes, model fit was evaluated using the chi-square statistic divided by degrees of freedom (χ²/df), the comparative fit index (CFI), the root mean square error of approximation (RMSEA), and the standardized root mean squared residual (SRMR). A CFI score above 0.95 signifies a good fit, while RMSEA and SRMR scores below 0.08 also indicate a good fit (Hu & Bentler, 1999). To compare the coefficients of negative and positive views, we compared two models. One model was unconstrained, with coefficients estimated freely. The second model was constrained so that the coefficient of the negative views was equal to the coefficient of the positive views. Coefficients would differ significantly if the unconstrained model fit significantly better than the constrained model.

Missing-value analysis revealed that 0.6%–10.8% of cases were missing for the various study variables. Applying Little’s (1988) statistical test to assess whether missing items were missing at random revealed that the data were missing completely at random, χ²(27) = 33.08, p = .19. Imputation of missing data was performed by replacing values with maximum-likelihood estimates when running models in AMOS.

Results

Table 1 presents the correlations between the study’s variables. AAQ-Loss and AARC-Losses were both related to lower RE. Most AAQ subscales were associated with AARC subscales; although some correlations were strong, most were moderate. AAQ-Loss showed moderate negative correlations with AAQ-Physical and AAQ-Growth. AARC-Losses and AARC-Gains did not correlate with each other.

Table 1.

Descriptive statistics for the study variables.

Variable	M/n	SD/%	1	2	3	4	5	6	7	8	9	10	11
1. Age	77.15	6.71	–
2. Gender (women)	95	60.4	0.025	–
3. Education	11.70	3.28	−0.083	−0.131	–
4. Family status (with partner)	80	50.9	−0.005	0.380**	0.008	–
5. Subjective health	2.86	1.15	0.180*	0.070	0.059	0.018	–
6. Mini-mental score	27.08	2.13	−0.054	−0.020	0.221**	0.017	0.100	–
7. AAQ-Loss	2.33	1.10	0.060	−0.044	−0.018	0.087	−0.275**	−0.193*	–
8. AAQ-Physical	3.33	1.14	−0.037	0.020	0.063	0.082	0.324**	−0.006	−0.322**	–
9. AAQ-Growth	4.28	0.89	−0.012	−0.002	−0.105	0.086	0.135	0.038	−0.217**	0.370**	–
10. AARC-Gains	4.53	0.62	−0.033	0.064	−0.102	0.058	−0.035	0.087	0.066	0.212**	0.331**	–
11. AARC-Losses	3.01	1.21	0.190*	0.037	−0.135	0.050	−0.225**	−0.232**	0.634**	−0.445**	−0.218**	0.018
12. RE	0.43	0.22	−0.136	−0.071	0.059	0.087	0.049	0.083	−0.192*	0.061	0.050	0.070	−0.192*

Note. AAQ = Attitudes toward Aging Questionnaire; AARC = awareness of age-related changes; RE = rehabilitation efficiency.

^* p < .05.

^** p < .01.

The model in which AAQ subscales predict RE showed an excellent fit to the data: χ²(34) = 25.86, p = .84, χ²/df = 0.76, CFI = 1.00, RMSEA = 0.00 (90% CI: .00, .04), SRMR = .05. AAQ-Loss predicted RE, such that an increase of 1 unit in AAQ-Loss predicted a decrease of 4% in RE. AAQ-Physical and AAQ-Growth did not predict RE. The covariates did not predict RE as well. The model explained 10.2% of the variance of RE. For model coefficients, see Table 2 and Figure 1.²

Table 2.

Attitudes toward Aging Questionnaire (AAQ) scores as predictors of rehabilitation efficiency (RE).

Variables	B	r	SE	β	p
Regression coefficients
AAQ-Loss → RE	−0.04		0.02	−0.19	.01
AAQ-Physical → RE	0.003		0.02	0.02	.85
AAQ-Growth → RE	0.007		0.02	0.03	.74
Age → RE	−0.003		0.003	−0.09	.26
Gender (women) → RE	−0.04		0.04	−0.09	.29
Education → RE	0.001		0.006	0.02	.85
Family status (with partner) → RE	−0.07		0.04	−0.15	.07
Subjective health → RE	0.000		0.02	0.000	.99
Condition (fractures) → RE	−0.07		0.04	−0.13	.11
Mini-mental score → RE	0.03		0.05	0.05	.50
Correlations
AAQ-Loss ↔ AAQ-Physical		−0.31			<.001
AAQ-Loss ↔ AAQ-Growth		−0.18			.02
AAQ-Physical ↔ AAQ-Growth		0.33			<.001
AAQ-Loss ↔ subjective health		−0.26			<.001
AAQ-Physical ↔ subjective health		0.29			<.001
Age ↔ subjective health		0.22			.003
Age ↔ condition (fractures)		0.22			.003
Gender (women) ↔ condition (fractures)		0.28			<.001
Gender (women) ↔ family status		−0.37			<.001
Mini-mental score ↔ condition (fractures)		0.14			.05
Mini-mental score ↔ education		0.22			.007

Figure 1.

Attitudes toward Aging Questionnaire (AAQ) scores as predictors of rehabilitation efficiency (RE). Values refer to standardized estimates. Values in bold italics refer to significant estimates at the level of p < .05. MMSE = Mini-Mental State Exam.

We compared the coefficients (AAQ-Loss vs AAQ-Physical and AAQ-Growth) by comparing two models. One model was unconstrained, with coefficients estimated freely. The second model was constrained so that the coefficient for AAQ-Loss was equal to the coefficients of AAQ-Physical and AAQ-Growth. Both the unconstrained and constrained models had a good fit (χ²[34] = 25.86, p = .84; and χ²[36] = 32.04, p = .66, respectively). However, the unconstrained model had a significantly better fit than the constrained model (Δχ²[2] = 6.20, p = .04).³

Therefore, as hypothesized, the effect of AAQ-Loss on RE is indeed significantly stronger than the effects of AAQ-Physical and AAQ-Growth on RE.

Because the AAQ included items referring to generalized VoA and others to personal views, we separated the AAQ-Loss items into those addressing personal VoA and those addressing generalized VoA.⁴ These two AAQ scores were initially entered into two separate path analysis models. The first model showed that a higher score on generalized AAQ-Loss items did not predict a lower score on RE, although the effect was close to being significant (B = −0.03, β = −0.15, p = .06). The second model demonstrated that a higher score on personal AAQ-Loss items predicted a lower score on RE (B = −0.03, β = −0.19, p = .014). Subsequently, each of the two AAQ-Loss scores (generalized and personal) was entered into the same path analysis to assess potential differences between them. Neither main effect was significant (i.e., neither generalized nor personal AAQ-Loss predicted RE), which possibly suggests a suppression effect, whereby one score suppressed the effect of the other on the dependent variable.

The model in which the AARC subscales predict RE also showed an excellent fit to the data: χ²(29) = 23.88, p = .74, χ²/df = 0.82, CFI = 1.00, RMSEA = 0.00 (90% CI = .00, .05), SRMR = .05. AARC-Losses predicted RE, such that an increase of 1 unit in AARC-Losses predicted a decrease of 3% in RE. AARC-Gains and covariates did not predict RE. The model explained 8.7% of the variance of RE. For model coefficients, see Table 3 and Figure 2.

Table 3.

Awareness of age-related change (AARC) scores as predictors of rehabilitation efficiency (RE).

Variables	B	R	SE	β	p
Regression coefficients
AARC-Losses → RE	−0.03		0.01	−0.17	.03
AARC-Gains → RE	−0.03		0.03	−0.09	.24
Age → RE	−0.003		0.003	−0.08	.34
Gender (women) → RE	−0.04		0.04	−0.08	.35
Education → RE	0.001		0.006	0.01	.89
Family status (with partner) → RE	−0.06		0.04	−0.14	.10
Subjective health → RE	0.004		0.02	0.02	.80
Condition (fractures) → RE	−0.07		0.04	−0.13	.12
Mini-mental score → RE	0.02		0.05	0.04	.62
Correlations
AARC-Losses ↔ subjective health		−0.27			<.001
Age ↔ subjective health		0.24			.002
Age ↔ condition (fractures)		0.21			.004
Gender (women) ↔ condition (fractures)		0.28			<.001
Gender (women) ↔ family status (with partner)		−0.37			.001
Mini-mental score ↔ condition (fractures)		0.14			.06
Mini-mental score ↔ education		0.22			.007

Note. RE = rehabilitation efficiency; AARC = awareness of age-related changes.

Figure 2.

Awareness of age-related change (AARC) scores as predictors of rehabilitation efficiency (RE). Values refer to standardized estimates; values in bold italics refer to significant estimates at the level of p < .05. MMSE = Mini-Mental State Exam.

Again, we compared an unconstrained model (in which effects were estimated freely) to the constrained model (in which the effects of AARC-Loss and AARC-Gains were constrained to be equal).⁵ There was no significant difference between the models (Δχ²[1] = 0.001, p = .96). This meant that although AARC-Loss was a significant predictor of RE while AARC-Gains was not, their coefficients were of similar magnitude (AARC-Loss was significant while AARC-Gains was not due to the smaller SE of the former).

Finally, we performed additional analyses comparing personal and generalized VoA by comparing AARC-Loss (personal) vs AAQ-Loss (the two generalized items). Neither main effect was significant (i.e., neither AARC-Loss nor generalized items from AAQ-Loss predicted RE), which possibly indicates a suppression effect, where one score reduces the impact of the other on the dependent variable.

Discussion

Our findings indicate that the negative facet of both generalized and personal VoA in two multi-dimensional and multi-directional VoA constructs significantly predicted rehabilitation outcomes in orthopedic and stroke patients, while the positive facet did not. Moreover, negative views had a significantly stronger effect on rehabilitation outcomes than positive views, but only for the AAQ subscales (not for the AARC subscales). In what follows, we will try to account for this direction of effect. However, we must first clarify that although AAQ and AARC differ conceptually, reflecting generalized versus personal VoA, their similar predictive effects on rehabilitation outcomes may be explained according to the stereotype embodiment theory (Levy, 2009), by which age-related beliefs, regardless of their original referent, become internalized over time and shape self-expectations, perceived control, and behavioral engagement. Thus, according to our findings, negative VoA, whether framed as generalized or as personal, may converge at the functional level.

Three theoretical frameworks converge on the stronger effects of negative VoA on health: the stereotype embodiment theory (Levy, 2009), the cognitive model of self-directed ageism (Henry et al., 2023), and the negativity bias theory (Rozin & Royzman, 2001). In the following, we will elaborate on how each model/theory contributes to clarifying the mechanisms linking negative VoA to worse rehabilitation outcomes.

Firstly, the combination of the stereotype embodiment theory (Levy, 2009) and the cognitive model of self-directed ageism (Henry et al., 2023) can account for our results. According to Levy (2009), negative age stereotypes that people are exposed to from childhood onward become internalized in older age, turning into a self-fulfilling prophecy expressed as self-directed ageism and as less engagement in preventive health behaviors. Accordingly, patients with negative age stereotypes in rehabilitation settings may attribute their medical condition to age and perceive it as less controllable and less curable; the expectation comes true and is manifest as less successful rehabilitation.

In addition, Henry et al. (2023) build on the above theory. According to their model, declines in executive cognitive functions are detrimental, leading to the dominance of negative VoA. Since this decline makes it harder for older people to learn new information, they rely more heavily on prior knowledge, such as ageist beliefs they have absorbed throughout their lives, which become internalized as self-directed ageism—exactly as Levy (2009) stated. Secondly, the same cognitive decline affects cognitive flexibility (Corbo et al., 2024), rendering it challenging to change these negative VoA with age. Accordingly, even exposure to positive VoA (i.e., new information) does not protect against the magnitude of the already existing knowledge of negative VoA. Thirdly, and according to Henry et al.’ s (2023) model, aging is accompanied by an increasing reliance on external cues in such a way that any ageist cue in the environment can awaken self-directed ageism, which is accompanied by self-doubt (Henry et al., 2024) that may negatively affect rehabilitation. Indeed, such environmental cues (e.g., hospital gowns, lying in a hospital room) may be quite common and trigger experiences of ageism, as ageism within hospital settings may be reflected in everyday inpatient care (Araújo et al., 2023).

Lastly, our results can be understood in light of the cognitive-emotional mechanism known as negativity bias, a broader principle that describes the heightened impact of negative information (relative to positive information of equal intensity) on cognition, affect, and behavior (Rozin & Royzman, 2001). Consequently, rehabilitation successes are likely to be overshadowed by even minor setbacks in the shaping of expectations and motivation. This principle also helps explain why negative VoA have such a strong effect, even though they are relatively less frequently reported than positive VoA, even among older adults coping with medical conditions (Bodner et al., 2023). This issue is important, as even the uncommon negative VoA, as reflected by low scores, suggests that a hospitalized patient may not be fully aware of these effects.

When we compared the strength of the effects of VoA on RE, negative views had a significantly stronger effect on rehabilitation outcomes than positive views, but only for the AAQ subscales, not for the AARC subscales. Actually, the unstandardized coefficients for both AARC subscales were of similar magnitude, but the effect for AARC-Gains was not significant because its standard error was high. In any case, consistent with a series of previous studies (Sabatini et al., 2020), negative views were significant predictors of physical health outcomes while positive views were not.

While our findings align with several studies that noted the stronger effect of negative VoA on various health outcomes (Brothers et al., 2017; Komalasari et al., 2025; Korkmaz Aslan et al., 2017; Sabatini et al., 2020), they contradict other studies that showed that positive VoA affects health more strongly, as occurred in Wurm and Schäfer’s study (2022) or in McLachlan et al. (2020), in which positive VoA predicted lower mortality among a healthy, community-dwelling sample. This shows that, in some cases, the effect of positive VoA can be stronger than that of negative VoA. However, it is worth considering that both Wurm and Schäfer’s (2022) and McLachlan et al.’s (2020) studies were conducted among healthy, community-dwelling individuals. This is markedly different from the current study, which focused on patients who experienced an acute decline, followed by hospitalization, and subsequently prolonged rehabilitation. The specific rehabilitation environment in which we chose to test the effects of VoA may further add to this difference, as suggested by Henry et al.’s model (2023). In this view, our sample may have experienced more self-directed ageism simply by being in a different context for a prolonged period, as is typical of rehabilitation settings, which last for 6–8 weeks. Consequently, we cannot rule out the possibility that, in some contexts, such as younger or healthier samples, positive VoA may have been a more robust predictor of subsequent health than negative VoA.

We also explored whether personal and generalized views would differ in their effects on rehabilitation outcomes. Findings were mixed. When testing personal and generalized items from AAQ-Loss, only personal items predicted lower RE. However, when we compared the effects of AARC-Losses and generalized items from AAQ-Loss, both constructs were non-significant, possibly due to suppression. Future studies should further explore whether personal and generalized VoA differ in their effects on physical health outcomes.

Several limitations of the present study should be acknowledged. First, the sample size was modest. Therefore, the study may have limited statistical power to detect small effects. Second, participants were recruited from two rehabilitation hospitals in Israel, raising questions about the cultural- and healthcare-system specificity of the results. The generalizability of the findings depends on replicating them in other contexts. Third, we relied on the short forms of the AAQ and AARC scales, and although validated, some subscales showed somewhat low reliability in our sample. Although low, the internal reliability values in the current study fall within the range reported in the literature on both AAQ and AARC across different samples and languages (Genç et al., 2024; Khongsirisombat et al., 2025; Kristoper et al., 2026; Laidlaw et al., 2018; Nosrati et al., 2023; Windsor & Wilton-Harding, 2025). In addition, when performing the analyses with revised subscales with improved reliability, the findings were similar. Future research should aim to replicate the findings in larger, more diverse clinical populations, use full versions of the measures, and employ longitudinal or experimental designs to better establish directionality. Moreover, future research with pre- and post-event (i.e., fracture or stroke) measurements can provide more insight into how acute health crises affect VoA.

In conclusion, our findings demonstrate that negative VoA exerts a stronger effect on rehabilitation outcomes relative to positive VoA among older adults recovering from hip fracture or stroke in a hospital rehabilitative setting, especially for the AAQ dimensions. These effects underscore the clinical relevance of addressing negative VoA in rehabilitation programs. Targeting such negative beliefs may represent a promising avenue for improving functional recovery and quality of life in geriatric patients. On the other hand, we do not refute the need to address different ways of not only ameliorating negative VoA but also promoting positive VoA (cf. Diehl et al., 2022).

Supplementary material

Supplementary material is available at The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences online.

Funding

This research was funded by the Israeli President’s Fellowship for Excellence and Scientific Innovation.

Conflicts of interest

None declared.

Data availability

The data that support the findings of this study are available from the corresponding author (A. Mansour) upon reasonable request. The study was not preregistered.