Abstract
Background
Hip fracture and cognitive impairment often coexist in later life, but whether cognitive disadvantage precedes the fracture-report index, worsens afterward, or both remains unclear. We examined pre- and post-index cognitive trajectories among older adults with reported incident hip fracture and matched controls.
Methods
We used Health and Retirement Study (HRS) data from 1996–2016. From 5,833 eligible US adults aged ≥ 65 years without baseline hip fracture, 543 incident hip fracture cases were matched to 1,538 controls using exact age/sex matching and propensity-score matching on sociodemographic, cognitive, and depressive-symptom measures. Incident hip fracture was reported at biennial interviews; cases were indexed at the age of first reported fracture, and matched controls were assigned the same index age. Global cognition was the primary out- come; mental status and memory were secondary outcomes. Event-centered piecewise linear mixed-effects models estimated pre-index slopes, index differences, and post-index slope differences, with covariate adjustment for residual imbalance. Sensitivity analyses examined raw-score outcomes, event-time assumptions, and post-index follow-up completeness.
Results
Of 558 eligible incident hip fracture cases, 543 were matched to 1,538 controls across 543 matched sets; the maximum absolute standardized mean difference was 0.055. For the primary outcome, global cognition was similar eight years before the index, lower by the index interview, and further separated eight years after the index. The post-index slope difference indicated additional worsening for global cognition (-0.022 (95% confidence interval [CI] -0.042 to -0.002)). Among secondary outcomes, mental status showed a similar adverse post-index slope difference (-0.030 (95% CI -0.052 to -0.007)), whereas memory did not (-0.009 (95% CI -0.028 to 0.009)).
Conclusions
In this matched longitudinal analysis of US adults aged 65 years or older, reported hip fracture marked a more adverse trajectory of cognitive aging. Cognitive vulnerability was evident before the fracture-report index, cognition was lower by the index interview, and additional post-index divergence was most apparent for global cognition and mental status.
Supplementary Information
The online version contains supplementary material available at 10.1186/s12877-026-07865-1.
Keywords: Hip fracture, Cognitive aging, Longitudinal trajectories, Health and Retirement Study, Propensity-score matching, Linear mixed-effects model
In amatched HRS cohort of US adults aged 65 years or older, cognitive disadvantage was evident before the fracture-report index.
Global cognition was lower at the index interview and showed additional post-index divergence versus matched controls.
Mental status showed a similar post-index pattern, whereas memory showed no clear post- index slope difference.
Findings support attention to cognitive vulnerability around hip-fracture care while recognizing the observational design and interval-based fracture ascertainment.
Introduction
Hip fracture is a major late-life health event whose consequences extend beyond the acute orthopedic injury. Mortality after hip fracture has been reported at approximately 10% at 1 month and 30% at 1 year [1]. Cognitive impairment is also common in this population: in a prospective cohort of patients aged 65 years or older with acute hip fracture, 54% had mild-to-moderate cognitive impairment, 27% died within 12 months, and 35% of those living at home before fracture were newly admitted to a nursing home [2]. Recovery frequently requires rehabilitation; a Cochrane review of multidisciplinary rehabilitation included 28 trials and 5,351 older participants, underscoring the functional and residential consequences of hip fracture [3]. These features make hip fracture a sentinel event in aging populations, potentially reflecting accumulated vulnerability while also introducing new risks for subsequent decline.
Cognitive health is central to this process. Cognitive impairment is common among older adults with hip fracture and is associated with mortality, nursing home admission, and poorer functional or mobility recovery [2, 4, 5]. At the same time, patients with cognitive impairment have often been excluded or insufficiently represented in hip fracture trials [6]. Much of the existing literature has therefore treated cognition either as a baseline prognostic factor among patients after fracture or as a short-term postoperative outcome. Less is known about how cognitive function evolves over the longer period surrounding hip fracture, particularly whether adverse cognitive patterns are already present before the fracture-report interval.
The relationship between hip fracture and cognitive decline is likely bidirectional. Cognitive impairment and dementia may increase fracture risk through falls, impaired judgment, comorbidity, osteoporosis, and treatment-related pathways [7, 8]. Conversely, fracture-related trauma, surgery, hospitalization, and delirium may plausibly contribute to subsequent cognitive worsening, especially among older adults with pre-existing physiologic or brain vulnerability [9–11]. This pattern is consistent with frailty and delirium vulnerability frameworks, in which baseline susceptibility and acute precipitating stressors jointly shape later health trajectories [12, 13]. Yet without repeated cognitive assessments before fracture, it remains difficult to determine whether post-fracture cognitive differences primarily reflect pre-existing vulnerability, additional worsening after the event, or both.
To address this gap, we used data from the Health and Retirement Study (HRS), a nationally representative longitudinal study with repeated cognitive assessments and established cognition resources [14, 15]. We examined event-centered associations and trajectory differences around the fracture-report interval, rather than estimating a discontinuity at a precisely dated fracture event. Specifically, we asked whether older adults who later reported hip fracture already showed faster cognitive decline before the index interview, whether cognition was lower by the index interview, and whether post-index cognitive change differed from that of comparable controls. By clarifying when cognitive differences emerge around hip fracture, this study may help identify older adults who need earlier cognitive assessment and more integrated post-fracture rehabilitation planning.
Methods
Data sources and study population
We conducted a secondary longitudinal cohort analysis using data from the HRS, a nationally representative biennial survey of adults aged 50 years and older in the United States. The HRS began in 1992 and collects repeated information on health, cognition, economic circumstances, and social factors; details of the study design and data collection procedures have been reported elsewhere [14].
We used HRS waves 3 through 13, corresponding to interviews conducted from 1996 to 2016, and wave 3 as the analytic baseline. Data were drawn from the RAND HRS Longitudinal File 2022 (Version 1) and the Harmonized HRS dataset developed by the Gateway to Global Aging Data. The RAND HRS file provided cognitive measures and demographic covariates, whereas the Harmonized HRS file provided harmonized health and fracture-related variables, including self-reported hip fracture.
The baseline analytic cohort included 5833 respondents who participated in wave 3, were aged 65 years, had no reported hip fracture at baseline, had baseline global cognition, had at least one follow-up global cognition assessment, and had complete baseline covariates required for matching and model adjustment. Incident hip fracture cases and no-hip-fracture controls were then assembled into the matched event-centered cohort described below. Sample selection and matched cohort construction are summarized in Fig. 1.
Fig. 1.
Sample selection and matched event-centered cohort construction. Legend. The analytic cohort was first defined using HRS eligibility and cognition criteria; incident hip fracture cases were then matched to no-hip-fracture controls. Note. Sequential exclusions and counts are shown in the flow diagram. Matching used exact integer baseline age and sex plus nearest-neighbor propensity-score matching with a target 3:1 control-to-case ratio, no replacement, and a 0.2-standard-deviation logit propensity-score caliper
The HRS was approved by the University of Michigan Institutional Review Board, and respondents provided informed consent at each wave. The present secondary analysis used de-identified public-use data.
Assessment of cognitive function
Cognitive function was assessed using the HRS core cognition battery, which is based on an adapted version of the Telephone Interview for Cognitive Status (TICS) and is available through HRS and RAND cognition resources [15, 16]. The primary outcome was global cognition, an HRS composite score summarizing overall cognitive performance. Secondary outcomes were mental status and memory component scores. Higher scores indicate better cognitive performance.
Memory was measured using total word recall, defined as the sum of immediate and delayed free recall of a 10-word list, with a range of 0–20. Mental status was measured using the HRS mental status summary score, with a range of 0–15, based on tasks including serial 7s, backward counting, object naming, date naming, and President/Vice-President naming. Global cognition ranged from 0–35 and combined the total recall and mental status indices. For comparability across reported outcomes, each outcome was standardized to a z-score using the wave 3 mean and standard deviation from the analytic cohort.
Assessment of hip fracture
Hip fracture was ascertained at each HRS interview from self-report of whether the respondent had fractured the hip since the previous interview. Incident hip fracture was defined as the first follow-up interview wave reporting hip fracture among respondents with no reported hip fracture at baseline. The fracture-report index was defined as the interview age at this first report. Because HRS ascertains hip fracture over an interval between interviews, the index identifies the reporting interview rather than the exact fracture date.
Covariates
Covariates were selected a priori based on prior evidence linking sociodemographic factors, cognitive function, depressive symptoms, frailty-related vulnerability, and hip fracture risk, and to support matching comparability [7, 8, 12, 15, 17]. All matching and model-adjustment covariates were measured at wave 3 baseline and treated as baseline covariates. These covariates included age in years, sex, race/ethnicity, education, marital status, depressive symptoms, baseline memory, and baseline mental status. Marital status was categorized as married/partnered, separated/divorced, or widowed/never married. Depressive symptoms were measured using the 8-item Center for Epidemiologic Studies Depression Scale (CES-D), with higher scores indicating more depressive symptoms.
Matched event-centered cohort
Incident hip fracture cases were matched to controls who did not report hip fracture during follow-up. Matching was exact on integer baseline age in years and sex and used nearest-neighbor propensity-score matching with a logistic propensity score, the average treatment effect in the treated estimand, a target 3:1 control-to-case ratio, no replacement, and a 0.2-standard-deviation caliper on the logit propensity score. The propensity-score model included race/ethnicity, education, baseline marital status, baseline CES-D, baseline memory, and baseline mental status.
For each matched set, the case index age was defined as the age at the first interview reporting hip fracture. Matched controls were assigned the index age of the case in the same matched set. Relative time was defined as interview age minus the assigned index age, so negative values represented years before the fracture-report or matched index and positive values represented years after the index. Covariate balance was assessed using absolute standardized mean differences (SMDs), with values below 0.10 interpreted as indicating adequate post-match balance, consistent with common propensity-score diagnostic practice [18].
Event-centered mixed-effects models
We fitted event-centered piecewise linear mixed-effects models separately for each standardized cognitive outcome. Pre-index time was defined as min(trel, 0) and post-index time as max(trel, 0), where trel was relative time in years. Models included hip fracture status, pre-index time, post- index time, the hip fracture by pre-index time interaction, and the hip fracture by post-index time interaction. Models adjusted for baseline age, sex, race/ethnicity, education, marital status, and CES-D score, and included participant-specific random intercepts and random slopes for relative time.
We summarized three event-centered parameters. The index difference was estimated by the hip fracture main-effect parameter and represented the adjusted between-group difference at the fracture-report index. The pre-index slope difference was estimated by the hip fracture by pre-index time interaction and represented the between-group difference in cognitive change before the index. The post-index slope difference was estimated by the hip fracture by post-index time interaction and represented the between-group difference in cognitive change after the index. The three event-centered parameters were the primary model estimates. Models were specified with participant-specific random intercepts and random slopes for event-centered time. When this specification failed to converge or produced a singular fit, the prespecified fallback model retained the participant-specific random intercept and omitted the random slope. Model-fitting diagnostics are reported in the Supplementary Appendix.
Supplementary and sensitivity analyses
We conducted supplementary analyses to examine the matching ratio, alternative event-centered model specifications, supportive raw-score results, and exploratory subgroup patterns. Matching-ratio sensitivity analyses repeated matching with target 1:3, 1:5, and 1:8 control-to-case ratios while retaining the same exact matching variables, propensity-score covariates, no-replacement rule, and caliper. To facilitate comparison across cognitive domains, the primary analyses used standardized cognitive outcomes. We repeated the primary models using raw cognitive scores on their original scales as supportive original-scale checks. Matched-design sensitivity analyses restricted relative time to −10 to +10 years and required cases to have at least two post-index observations while retaining controls from the corresponding matched sets. Exploratory subgroup analyses were limited to post-index slope differences because the subgroup question concerned whether post-index cognitive divergence varied by sex, baseline age group, or education. We estimated stratum-specific post-index slope differences descriptively and used P values for interaction to evaluate between-stratum heterogeneity. Detailed supplementary estimates, subgroup figures, and model diagnostics are reported in the Supplementary Appendix.
Analyses were conducted using R version 4.5.3. All P-values were two-sided, and statistical significance was evaluated at a two-sided alpha level of 0.05.
Results
Analytic sample and matched cohort
Among 17,991 HRS wave 3 baseline participants, sequential eligibility criteria yielded 5,833 respondents in the analytic cohort (Fig. 1). Compared with excluded wave 3 participants, included respondents were older by design (mean age 75.2 (6.5) vs 62.1 (10.4) years), had a similar male distribution (2437 (41.8%) vs 5042 (41.5%)), and were more often White/Caucasian (5027 (86.2%) vs 9765 (80.3%); Table S1). Within the analytic cohort, 558 respondents reported incident hip fracture and 5,275 did not report hip fracture during follow-up; their pre-matching baseline characteristics are reported descriptively in Table S2. Exact age/sex plus propensity-score matching matched 543 incident hip fracture cases to 1,538 controls across 543 matched sets; 15 incident cases remained unmatched (Table S3). At baseline, matched hip fracture cases and controls were similar in age, sex, race/ethnicity, education, marital status, depressive symptoms, and cognitive scores (Table 1). The maximum post-match absolute SMD was 0.055; detailed balance diagnostics are provided in Table S4 and Figure S1.
Table 1.
Baseline characteristics of the matched cohort by incident hip fracture status
| Characteristic | No hip fracture | Hip fracture | Absolute SMD after matching |
|---|---|---|---|
| Baseline age, years | 75.95 (5.96) | 76.16 (6.10) | 0.035 |
| Sex | |||
| Male | 330 (21.5%) | 116 (21.4%) | 0.002 |
| Female | 1208 (78.5%) | 427 (78.6%) | 0.002 |
| Race/ethnicity | |||
| White/Caucasian | 1375 (89.4%) | 492 (90.6%) | 0.040 |
| Black/African American | 151 (9.8%) | 45 (8.3%) | 0.053 |
| Other | 12 (0.8%) | 6 (1.1%) | 0.034 |
| Education | |||
| Less than high school | 557 (36.2%) | 201 (37.0%) | 0.017 |
| GED | 29 (1.9%) | 11 (2.0%) | 0.010 |
| High school graduate | 520 (33.8%) | 161 (29.7%) | 0.089 |
| Some college | 270 (17.6%) | 106 (19.5%) | 0.051 |
| College or above | 162 (10.5%) | 64 (11.8%) | 0.040 |
| Marital status | |||
| Married/Partnered | 797 (51.8%) | 285 (52.5%) | 0.013 |
| Separated/Divorced | 68 (4.4%) | 18 (3.3%) | 0.057 |
| Widowed/Never married | 673 (43.8%) | 240 (44.2%) | 0.009 |
| Baseline CES-D score | 1.41 (1.85) | 1.37 (1.77) | 0.024 |
| Baseline memory | 8.97 (3.58) | 8.80 (3.69) | 0.047 |
| Baseline mental status | 12.44 (2.40) | 12.49 (2.41) | 0.022 |
| Baseline global cognition | 21.41 (5.13) | 21.29 (5.23) | 0.023 |
Values summarize baseline characteristics in the matched event-centered cohort by incident hip fracture status
Continuous variables are shown as mean (SD), and categorical variables are shown as n (%). Age is in years. SD indicates standard deviation; CES-D, Center for Epidemiologic Studies Depression scale; GED, General Educational Development; SMD, standardized mean difference. Matching balance was assessed using absolute standardized mean differences; values below 0.10 were interpreted as indicating acceptable post-match balance. P-values were not used to judge matching quality
Event-centered cognitive trajectories
For the primary outcome, global cognition was similar between matched groups eight years before the index, lower in the hip fracture group by the index interview, and more separated eight years after the index. Secondary outcomes showed a similar pattern for mental status and a smaller post-index separation for memory (Fig. 2).
Fig. 2.
Cognitive trajectories before and after the fracture-report index. Legend. The figure shows model-based predicted standardized cognitive scores for global cognition, mental status, and memory. Note. Matched controls were assigned the index age of their matched case; the index corresponds to the first interview reporting hip fracture for cases
At eight years before the index, predicted between-group differences were small across reported outcomes (range −0.032 to 0.041 z-score units; Table S5). By the index interview, the hip fracture group had lower predicted cognition across outcomes (range −0.229 to −0.202), and differences at eight years after the index were largest for global cognition and mental status (Table S5).
For global cognition, the predicted between-group difference changed from 0.005 eight years before the index to −0.229 at the index and −0.407 eight years after the index. For mental status, the corresponding differences were 0.041, −0.212, and −0.451.
Memory showed a more limited post-index pattern than global cognition and mental status (for complete predicted contrasts at −8, 0, and +8 years, see Table S5; for formal event-centered parameter estimates, see Table 2 and Fig. 3).
Table 2.
Key estimates from matched event-centered linear mixed-effects models for reported cognitive outcomes
| Outcome | Index difference | Pre-index slope difference | Post-index slope difference |
|---|---|---|---|
| Global cognition |
−0.229 (−0.331, −0.127) p = < 0.001 |
−0.029 (−0.039, −0.020) p = < 0.001 |
−0.022 (−0.042, −0.002) p = 0.028 |
| Mental status |
−0.212 (−0.327, −0.098) p = < 0.001 |
−0.032 (−0.042, −0.021) p = < 0.001 |
−0.030 (−0.052, −0.007) p = 0.009 |
| Memory |
−0.202 (−0.284, −0.121) p = < 0.001 |
−0.021 (−0.030, −0.013) p = < 0.001 |
−0.009 (−0.028, 0.009) p = 0.315 |
The table provides the numeric estimates corresponding to Fig. 3
Each cell shows the fixed-effect estimate with 95% confidence interval and p-value. Index difference compares hip fracture cases with matched controls at the assigned index interview. Pre-index and post-index slope differences are hip fracture by relative-time interaction terms
Fig. 3.
Between-group differences in cognitive level and change around the fracture-report index. Legend. The figure summarizes the adjusted index difference, pre-index slope difference, and post-index slope difference across the reported cognitive outcomes. Note. Points are estimates and horizontal bars are 95% confidence intervals
Event-centered model estimates by cognitive outcome
For the primary outcome, global cognition, the hip fracture group had lower cognition at the index interview (−0.229, 95% CI −0.331 to −0.127; p=<0.001), faster pre-index decline (−0.029, 95% CI −0.039 to −0.020; p=<0.001), and additional post-index decline (−0.022, 95% CI −0.042 to −0.002; p=0.028; see Table 2 and Fig. 3).
Among secondary outcomes, mental status also showed lower cognition at the index interview (−0.212, 95% CI −0.327 to −0.098; p=<0.001), faster pre-index decline (−0.032, 95% CI −0.042 to −0.021; p=<0.001), and additional post-index decline (−0.030, 95% CI −0.052 to −0.007; p=0.009). Memory showed lower cognition at the index interview (−0.202, 95% CI −0.284 to −0.121; p=<0.001), faster pre-index decline (−0.021, 95% CI −0.030 to −0.013; p=<0.001), but no clear post-index slope difference (−0.009, 95% CI −0.028 to 0.009; p=0.315). Full fixed-effect coefficients are reported in Table S11.
Sensitivity analyses of the matched design
Sensitivity analyses assessed whether the post-index pattern was similar under alternative matching ratios or matched-design specifications. With target 1:5 and 1:8 matching, post-index slope differences remained adverse for global cognition and mental status, whereas memory confidence intervals continued to include the null (Table S7). Supportive raw-score analyses on the original cognitive score scales were directionally consistent with the standardized models: post-index slope differences remained adverse for global cognition and mental status, whereas the memory estimate remained compatible with no clear difference (Table S8). Restricting relative time to −10 to +10 years retained the post-index pattern for global cognition and mental status. Requiring hip fracture cases to have at least two post-index cognitive observations attenuated post-index estimates, and no post-index slope difference remained statistically significant (Table S9). Model diagnostics are summarized in Table S6.
Exploratory subgroup analyses
In exploratory subgroup analyses, evidence for heterogeneity was limited (Table S10; Figure S2–S4). The clearest signal was baseline age group for mental status: the post-index slope difference was adverse among participants younger than 80 years (−0.046, 95% CI −0.070 to −0.022; p=<0.001) but not among those aged 80 years or older (0.054, 95% CI −0.008 to 0.117; p=0.087; P for interaction <0.001). Global cognition showed a weaker age pattern (age <80 years: −0.030, 95% CI −0.052 to −0.008; p=0.006; age 80 years: 0.016, 95% CI −0.035 to 0.066; p=0.545; P for interaction 0.073), and memory did not show age-related heterogeneity (P for interaction 0.729). P values for interaction did not support consistent heterogeneity by sex (global cognition p=0.334; mental status p=0.924; memory p=0.281) or education group (global cognition p=0.586; mental status p=0.455; memory p=0.601).
Discussion
Principal findings
In this matched event-centered longitudinal analysis, three findings emerged. First, participants who later reported hip fracture had faster pre-index decline in global cognition, the primary outcome, and in the secondary outcomes. Second, cognition was lower by the index interview. Third, additional post-index divergence was most apparent for global cognition and mental status, whereas memory showed less clear post-index separation. These findings suggest that hip fracture is closely linked with cognitive vulnerability already present before the reporting interval, with possible additional post-index divergence in selected cognitive measures.
Comparison with previous studies
Our findings are broadly consistent with previous evidence linking cognitive impairment and dementia to hip fracture risk, adverse post-fracture outcomes, institutionalization, mortality, and poorer functional recovery [2, 4, 5, 7, 8]. Prior studies have also highlighted postoperative delirium, perioperative complications, and heterogeneous post-fracture cognitive trajectories after hip fracture [9, 10, 19–21]. However, much of this literature has enrolled patients during hospitalization, after surgery, or in the early post-fracture period. As a result, cognition has often been treated as a baseline prognostic factor after fracture or as a short-term postoperative outcome, rather than as a dynamic process that may already have been changing before fracture was reported.
This study adds a temporal dimension to this evidence by aligning repeated cognitive assessments before and after the fracture-report index [14, 15]. This design allowed us to ask when cognitive disadvantage emerged. The finding that predicted cognitive differences were small in the earlier pre-index period but had become apparent by the fracture-report index suggests that poorer cognition after hip fracture cannot be interpreted solely because of the fracture episode. Rather, an adverse cognitive trajectory may already have been developing before fracture was reported, supporting the interpretation of hip fracture as a marker of accumulated vulnerability. The post-index divergence observed for global cognition and mental status raises the possibility of subsequent worsening, but this interpretation requires caution because hip fracture was ascertained at biennial interviews and post-index estimates were sensitive to follow-up completeness.
Possible mechanisms
The faster pre-index cognitive decline among participants who later reported hip fracture is unlikely to reflect a direct effect of the fracture itself. Rather, it may indicate that hip fracture and cognitive decline are partly different manifestations of a shared vulnerability process. Cognitive decline, depressive symptoms, impaired gait and balance, executive dysfunction, higher fall risk, multimorbidity, polypharmacy, frailty, and reduced physiologic reserve may jointly increase both the likelihood of hip fracture and the rate of cognitive deterioration [7, 8, 12]. In this sense, hip fracture may function as a sentinel event that marks an already adverse trajectory of cognitive and physiologic aging.
After the fracture-report index, the fracture episode may also be followed by additional cognitive worsening among vulnerable older adults. Trauma, hospitalization, surgery, inflammatory responses, pain, sleep disruption, immobility, sedative or analgesic exposure, postoperative delirium, deconditioning, and insufficient rehabilitation may interact with pre-existing vulnerability and reduce the capacity for cognitive and functional recovery. These pathways are consistent with frailty and delirium vulnerability frameworks, in which acute stressors may have disproportionate and persistent consequences when baseline physiologic and cognitive reserve is already limited [13, 22–25].
The domain-specific pattern should be interpreted cautiously. In the primary analysis, additional post-index divergence was clearest for global cognition and mental status, whereas memory showed less consistent evidence of post-index acceleration. One possible interpretation is that orientation-, attention-, and executive-related functions captured by the HRS mental status measure may be particularly sensitive to acute illness, delirium, hospitalization, immobility, and rehabilitation disruption. However, global cognition is a composite measure that includes mental status; therefore, the post-index difference in global cognition may be driven in part, and perhaps substantially, by the mental status component, and the parallel findings for these two outcomes should not be interpreted as fully independent domain-specific effects. Because HRS does not provide exact fracture dates, detailed perioperative information, delirium assessments, medication exposure, or rehabilitation intensity, these mechanisms should be viewed as plausible explanatory pathways rather than processes directly tested in the present study.
Strengths and limitations
This study has several strengths. The HRS is a nationally representative longitudinal cohort of older US adults, allowing repeated cognitive assessments over years before and after the fracture-report index. The matched event-centered design assigned controls the index age of their matched case and improved comparability between incident hip fracture cases and no-hip-fracture controls; post-match balance was strong, with a maximum absolute SMD of 0.055. The analysis also prespecified global cognition as the primary outcome, retained mental status and memory as secondary outcomes, and examined robustness through raw-score, event-time, matching-ratio, and post-index follow-up sensitivity analyses.
Several limitations should be considered when interpreting the findings. Hip fracture in the HRS was ascertained by self- or proxy-report at biennial interviews, so the fracture-report index was the first interview at which hip fracture was reported rather than the exact fracture date. The index difference may therefore include cognitive vulnerability already present before fracture and early post-fracture changes occurring within the reporting interval. In addition, controls were participants who did not report hip fracture during observed follow-up, so the primary comparison was with persistent no-hip-fracture controls rather than risk-set controls sampled among participants still at risk at the index age.
Residual confounding and health selection remain possible despite matching. Unmeasured or in- completely measured differences in frailty severity, comorbidity, function, prior falls, medication exposure, osteoporosis severity, and health-care access may have contributed to both hip fracture risk and cognitive decline. The HRS also lacks detailed peri-fracture clinical information, including fracture type, surgery, anesthesia, delirium, complications, medication exposure, and rehabilitation intensity. Finally, death, attrition, and selective follow-up may affect post-index estimates; the sensitivity analysis requiring at least two post-index observations attenuated post-index slope differences and selected a longer-observed subset. Generalizability may also differ in health systems or populations with different fracture care, rehabilitation, long-term care access, or sociodemographic composition.
Clinically, the findings support attention to cognitive vulnerability in older adults before and after hip fracture. Pre-fracture cognitive trajectory information may help identify vulnerable patients, and post-fracture cognitive follow-up may be useful, while these observational findings should not be interpreted as evidence that fracture care itself caused the observed cognitive differences.
Future research
Future research should stay close to the event-timing and comparison-group questions raised by this study. Risk-set matching or incidence-density sampling could better align controls with cases at the time a fracture is reported. Analyses that examine alternative index definitions or midpoint assumptions could clarify how interval-reported hip fracture timing affects event-centered estimates. Linkage of longitudinal survey data with hospitalization, surgery, delirium, rehabilitation, medication, mortality, and long-term care records would help distinguish pre-existing vulnerability from peri-fracture factors that may contribute to post-index cognitive change.
Conclusion
In this matched longitudinal analysis of US adults aged 65 years or older, reported hip fracture marked a more adverse trajectory of cognitive aging. Cognitive vulnerability was already evident before the fracture-report index, and cognitive performance was lower by the index interview. The index-interview difference should therefore be interpreted as a difference already present by the reporting visit, potentially reflecting both prior vulnerability and early changes within the reporting interval. Additional adverse post-index change was most apparent for global cognition and mental status, whereas memory showed a less clear post-index pattern. Because post-index estimates were sensitive to follow-up completeness, the post-index pattern should be interpreted cautiously.
Supplementary Information
Acknowledgements
We thank the HRS participants and research team for making the data available for research.
Clinical trial number
Not applicable.
Abbreviations
- CES-D
Center for Epidemiologic Studies Depression Scale
- CI
Confidence interval
- GED
General Educational Development
- HRS
Health and Retirement Study
- LMM
Linear mixed-effects model
- RAND
RAND Corporation
- SD
Standard deviation
- SMD
Standardized mean difference
- TICS
Telephone Interview for Cognitive Status
- US
United States
Authors’ contributions
Study design: H.H. Data cleaning and analysis: H.H., F.D. Data interpretation: All authors. Preparation of manuscript: H.H. B.C., and Z.C. Manuscript review: All authors. The authors read and approved the final manuscript.
Funding
This study was supported by the National Natural Science Foundation of China (No. 82472566) and the National Clinical Key Specialty Construction Program of China. The funders had no role in the study design, data analysis, data interpretation, manuscript preparation, or decision to submit the manuscript for publication.
Data availability
The datasets analyzed in this study are available in the Health and Retirement Study repository: https://hrsdata.isr.umich.edu/.
Declarations
Ethics approval and consent to participate
This study was a secondary analysis of de-identified public-use data from the HRS. The original HRS was approved by the Institutional Review Board at the University of Michigan and was conducted in accordance with relevant guidelines and regulations. Written informed consent was obtained from participants in the original study.
Consent for publication
Not applicable.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Contributor Information
Haiyan Hu, Email: hyan87@gmail.com.
Bo Chen, Email: chenbo71@sina.com.
References
- 1.Downey C, Kelly M, Quinlan JF. Changing trends in the mortality rate at 1-year post hip fracture - a systematic review. World J Orthop. 2019;10(3):166–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Schaller F, Sidelnikov E, Theiler R, Egli A, Staehelin HB, Dick W, et al. Mild to moderate cognitive impairment is a major risk factor for mortality and nursing home admission in the first year after hip fracture. Bone. 2012;51(3):347–52. [DOI] [PubMed] [Google Scholar]
- 3.Handoll HH, Cameron ID, Mak JC, Panagoda CE, Finnegan TP. Multidisciplinary rehabilitation for older people with hip fractures. Cochrane Database Syst Rev. 2021;11(11):CD007125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Jones CA, Jhangri GS, Feeny DH, Beaupre LA. Cognitive status at hospital admission: postoperative trajectory of functional recovery for hip fracture. J Gerontol A Biol Sci Med Sci. 2016;72(1):61–7. [DOI] [PubMed] [Google Scholar]
- 5.Ariza-Vega P, Lozano-Lozano M, Olmedo-Requena R, Martín-Martín L, Jiménez-Moleón JJ. Influence of cognitive impairment on mobility recovery of patients with hip fracture. Am J Phys Med Rehabil. 2017;96(2):109–15. [DOI] [PubMed] [Google Scholar]
- 6.Mundi S, Chaudhry H, Bhandari M. Systematic review on the inclusion of patients with cognitive impairment in hip fracture trials: a missed opportunity? Can J Surg. 2014;57(4):E141–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Jeon JH, Park JH, Oh C, Chung JK, Song JY, Kim S, et al. Dementia is associated with an increased risk of hip fractures: a nationwide analysis in Korea. J Clin Neurol. 2019;15(2):243. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Friedman SM, Menzies IB, Bukata SV, Mendelson DA, Kates SL. Dementia and hip fractures. Geriatr Orthop Surg Rehabil. 2010;1(2):52–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Krogseth M, Watne LO, Juliebø V, Skovlund E, Engedal K, Frihagen F, et al. Delirium is a risk factor for further cognitive decline in cognitively impaired hip fracture patients. Arch Gerontol Geriatr. 2016;64:38–44. [DOI] [PubMed] [Google Scholar]
- 10.Umoh ME, Sharma A, Leoutsakos JMS, Lyketsos CG, Inouye SK, Marcantonio ER, et al. Cognitive outcomes after hip fracture surgery: the association of postoperative delirium on previously cognitively normal older adults. Am J Geriatr Psychiatry. 2026;34(6):857–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Alam A, Hana Z, Jin Z, Suen KC, Ma D. Surgery, neuroinflammation and cognitive impairment. EBioMedicine. 2018;37:547–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Clegg A, Young J, Iliffe S, Rikkert MO, Rockwood K. Frailty in elderly people. Lancet. 2013;381(9868):752–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Inouye SK. Predisposing and precipitating factors for delirium in hospitalized older patients. Dement Geriatr Cogn Disord. 1999;10(5):393–400. [DOI] [PubMed] [Google Scholar]
- 14.Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. Cohort profile: the health and retirement study (HRS). Int J Epidemiol. 2014;43(2):576–85. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Crimmins EM, Kim JK, Langa KM, Weir DR. Assessment of cognition using surveys and neuropsychological assessment: the health and retirement study and the aging, demographics, and memory study. J Gerontol B Psychol Sci Soc Sci. 2011;66B(Supplement 1):i162–71. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.McCammon RJ, Fisher GG, Hassan H, Faul JD, Rodgers WL, Weir DR. Imputation of Cognitive Functioning Measures: 1992–2020. Data Description; 2023.
- 17.Dang L, Dong L, Mezuk B. Shades of blue and gray: a comparison of the center for epidemiologic studies depression scale and the composite international diagnostic interview for assessment of depression syndrome in later life. Gerontologist. 2019;60(4):e242–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Stuart EA. Matching methods for causal inference: A review and a look forward. Stat Sci. 2010;25(1):1–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Beishuizen SJE, van Munster BC, de Jonghe A, Hanna AA, Buurman BM, de Rooij SE. Distinct cognitive trajectories in the first year after hip fracture. J Am Geriatr Soc. 2017;65(5):1034–42. [DOI] [PubMed] [Google Scholar]
- 20.Oughli HA, Chen G, Miller JP, Nicol G, Butters MA, Avidan M, et al. Cognitive improvement in older adults in the year after hip fracture: implications for brain resilience in advanced aging. Am J Geriatr Psychiatry. 2018;26(11):1119–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Min D, Lee HS, Shin M. Consequences of fall-induced hip fractures on cognitive function, physical activity, and mortality: Korean longitudinal study of aging 2006–2016. Injury. 2021;52(4):933–40. [DOI] [PubMed] [Google Scholar]
- 22.Chen Y, Liang S, Wu H, Deng S, Wang F, Lunzhu C, et al. Postoperative delirium in geriatric patients with hip fractures. Front Aging Neurosci. 2022;14:1068278. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Subramaniyan S, Terrando N. Neuroinflammation and Perioperative Neurocognitive Disorders. Anesth Analg. 2019;128(4):781–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Fong TG, Vasunilashorn SM, Ngo L, Libermann TA, Dillon ST, Schmitt EM, et al. Association of plasma neurofilament light with postoperative delirium. Ann Neurol. 2020;88(5):984–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Umoh ME, Yan H, Leoutsakos JM, Lewis A, Marcantonio ER, Lyketsos CG, et al. Preoperative neurofilament light associated with postoperative delirium in hip fracture repair patients without dementia. J Acad Consult Liaison Psychiatry. 2025;66(1):3–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets analyzed in this study are available in the Health and Retirement Study repository: https://hrsdata.isr.umich.edu/.