Introduction

Both in Western and Asian countries, hip fractures are common among older adults and are associated with a loss of physical function, increased dependence, and a poor quality of life, leading to a major burden on the healthcare system (Fischer et al., 2019; Peeters et al., 2016; Wantonoro et al., 2022). Diabetes mellitus (DM) is a risk factor for osteoporosis, which is a crucial risk factor for hip fracture (Gani et al., 2020). A previous study found that older adults with hip fracture and DM had poorer recovery outcomes after surgery than those without DM (Tebé et al., 2019). One study found 63%–68% of older adults with both hip fracture and DM were affected by abnormal sleep duration during the 6 months after hospital discharge (Kuo et al., 2016). Moreover, those with abnormal sleep duration (≤ 6 hours per day or ≥ 9 hours per day) had worse femoral muscle recovery on the fractured side and poorer blood glucose control than those with normal sleep duration (6.1–8.9 hours per day) during recovery from hip fracture (Kuo et al., 2016). However, the factors associated with abnormal sleep duration in older adults with hip fracture and DM remain uncertain.

Good sleep is a basic human need. Prior studies have suggested both short and long sleep durations as being associated with a range of adverse health outcomes (Itani et al., 2017; Jike et al., 2018). A systematic review found that, compared with normal sleep duration (6–8 hours/night), both short (4–5.9 hours/night) and long (≥ 8–9 hours/night) sleep durations in adult participants were significantly associated with cardiovascular-related diseases, including cardiovascular disease, stroke, coronary heart disease, and obesity (Itani et al., 2017; Jike et al., 2018). In addition, both short and long sleep durations can increase the overall risk of cardiovascular disease mortality (Krittanawong et al., 2019).

Several studies have explored the factors associated with abnormal sleep in older adults. A prior systematic review highlighted female gender, depressed mood, and physical illness as general risk factors for abnormal sleep in older adults. However, specific physiological pathways have yet to be established (Smagula, Stone, et al., 2016). Gulia and Kumar (2018) found that age-related medical problems such as hypertension, DM, renal failure, respiratory diseases, and anxiety are associated with abnormal sleep in older people. Yu et al. (2018) revealed social isolation to be a predictor of abnormal sleep in older people. In addition, the predictive factors of abnormal sleep in patients with DM have been investigated in several studies. The results of one prior study indicate that women, smokers, unemployed individuals, insulin users, and patients with uncontrolled diabetes face a significantly higher risk of abnormal sleep (Barakat et al., 2019). In addition, a systematic review found that multiple factors may contribute to abnormal sleep in patients with DM, including discomfort or pain associated with peripheral neuropathy, restless leg syndrome, periodic limb movements, and rapid changes in blood glucose levels during the night that lead to hypoglycemic and hyperglycemic episodes and nocturia (Khandelwal et al., 2017).

As described above, prior studies have reported that certain factors are associated with abnormal sleep in older adults or patients with DM. However, little is known whether hip-fracture-related factors other than those related to DM also predict abnormal sleep in older adults with both hip fracture and DM. The aim of this study was to assess whether demographic factors (age, gender, and comorbidities), hip-fracture-related factors (diagnosis and surgical methods), and DM-related factors (duration of DM, methods of DM control, diabetic peripheral neuropathy [DPN], and diabetic peripheral vascular disease) predicted abnormal sleep duration in hip-fractured older adults with DM within the 6-month period after hospital discharge. The results of several studies indicate old age, being female, and chronic diseases correlate with poor sleep in older people, although the related mechanisms have yet to be identified (Gulia & Kumar, 2018; Smagula, Stone, et al., 2016). A previous study found that DM complications negatively affect sleep (Khandelwal et al., 2017). Therefore, in this study, older age, being female, and having a larger number of comorbidities are hypothesized as predictors of abnormal sleep duration (Smagula, Stone, et al., 2016). In addition, we hypothesize that older adults who have a diagnosis of intertrochanteric fracture, receive open/closed reduction with internal fixation (ORIF/CRIF), have a long duration of DM, control their blood sugar with insulin, or have a diagnosis of DPN or diabetic peripheral vascular disease are at an increased risk of abnormal sleep duration after hospital discharge (Khandelwal et al., 2017).

Methods Design Current study

This longitudinal study used secondary data from a randomized controlled trial related to the effects of a diabetes-specific care model on older adults with hip fracture.

Parent study

The parent study was a randomized controlled trial conducted from January 1, 2010, to December 31, 2014 with a 24-month follow-up. The aim of the parent study was to develop a well-conceived and feasible protocol for hospital discharge and subacute care for older adults with hip fracture and diabetes.

Participants

In the parent study, the inclusion criteria were as follows: (a) at least 60 years old, (b) diagnosed with diabetes, and (c) admitted to the emergency room for hip fracture. One hundred seventy-six participants were recruited and randomly assigned to the experimental (n = 88) and control (n = 88) groups.

To avoid the effects of the intervention, we used only the data from the control group (n = 88) in this study to explore the targeted predictors (age, gender, comorbidities, diagnosis, surgical methods, duration of diabetes, methods of diabetes control, DPN, and diabetes-related peripheral vascular disease) of abnormal sleep duration. One of the participants died (first month after discharge), 10 refused to continue to participate, and 22 had incomplete data. Sleep duration was measured using a body monitoring system. Thus, data from 55 participants were available for analysis. To capture comprehensive sleep conditions from the participants, only data from a complete 24-hour recording were analyzed, including the sleep hours from at least one time point of the three measurements. We applied G*Power with a logistic regression model to calculate posterior power using the current adjusted odds ratio generalized estimating equations (GEE) model parameters (Lu et al., 2013). During the sixth month after surgery, older adults with hip fractures had better recovery and sleep conditions (Kuo et al., 2016; Shyu et al., 2004). In addition, because a correlation between DM complications and sleep conditions is supported in the literature (Khandelwal et al., 2017), the abnormal sleep duration incident rate was used in the sixth month after discharge together with predictive factors, including DPN and diabetes-related peripheral vascular disease, to calculate posterior power. This analysis showed that the sample size of 55 used in this study attained 91% and 95% power when, respectively, DPN (adjusted OR = 9.6) and diabetes-related peripheral vascular disease (adjusted OR = 15.62) were used to calculate posterior power.

The complete (n = 55) and incomplete (n = 33) sets of participant data were compared, with no significant differences found between them. The data included age (p = .90), gender (p = .60), education (p = .97), comorbidities (p = .80), diagnosis (p = .72), surgical methods (p = .09), duration of DM (p = .69), and methods of DM control (p = .87).

Data Collection

In the parent study, the participants were recruited from a medical center. Demographics and clinical data (i.e., age, gender, education, comorbidity, diagnosis, surgical methods, duration of DM, and methods of DM control) were assessed once before hospital discharge. The incidences of DPN and DM-related peripheral vascular disease were collected before discharge and at 1, 3, 6, 12, 18, and 24 months after discharge. Sleep duration was measured at 1, 3, 6, 12, 18, and 24 months after discharge. All postdischarge data were collected by two research nurses during home visits. Because the predictors of abnormal sleep duration during only the first 6 months after hospital discharge were examined in this study, only the data from 1, 3, and 6 months after discharge were used in the analysis.

Ethical Considerations

This study was approved for human subject testing by the study hospital (103-6642C) before the study began, and written informed consent was obtained from all participants. The participants were informed of their right to withdraw at any time during the study and were given a copy of the signed consent form.

Data Analysis

Means and percentages were used to describe the sample. The GEE approach was used to assess the relationship between the predictors (age, gender, comorbidities, diagnosis, surgical methods, duration of diabetes, methods of diabetes control, DPN, and diabetes-related peripheral vascular disease) and the binary outcome variable (abnormal sleep duration vs. normal sleep duration) in repeated measures over time. This approach accounted for possible correlations in the repeated measurements performed over time and allowed differences to be examined at different time points (Liang & Zeger, 1986, 1993). We adjusted for the major confounding factors expected to affect these relationships (i.e., time, education, depression symptoms, and physical pain; Andersen et al., 2018; Smagula, Stone, et al., 2016) and presented the multivariate model in the results. The IBM SPSS Statistics 19.0 (IBM Inc., Armonk, NY, USA) was used for statistical analysis.

Measurements Fracture-related factors

Information on fracture-related factors, which relate to the nature and treatment of hip fractures (e.g., diagnoses and surgical methods), were collected from medical charts after the participants agreed to participate.

Diabetes-related factors

Information on DM duration, methods used to control DM, and diabetes-related peripheral vascular disease was collected, respectively, by asking the following simple questions: “How many years have you been diagnosed with diabetes?”; “What doctor-prescribed method(s) do you use to control your diabetes?” (response options: diet control, oral medication control, insulin control, and oral medication plus insulin control); and “Have you been diagnosed with a diabetes-related peripheral vascular disease?” (with a “yes” response coded as 1). DPN was measured using the Michigan Neuropathy Screening Instrument (Pop-Busui et al., 2017). Research nurses inspected each foot for deformities, dry skin, calluses, infections, fissures, and ulcerations. Each foot exhibiting an abnormal appearance was scored as 1. A reflex hammer was used to examine the Achilles tendon reflex. A research nurse lightly held each participant's distal sole and percussed the Achilles tendon. Achieving optimal stretch earned a score of 0. If the reflex response was absent, the participant was instructed to perform the Jendrassik maneuver (hooking the fingers together and pulling), after which the research nurse percussed the Achilles tendon again. If the reflex response was still absent, a score of 1 was assigned. Vibration sensation was assessed using a 128-Hz tuning fork, with a score of 1 assigned if the sensation was absent in the target foot. Total possible scores ranged from 0 to 10, and a score of ≥ 2.5 was defined as abnormal (coded as 1; Herman et al., 2012). Good sensitivity (80%) and specificity (95%) for the Michigan Neuropathy Screening Instrument have been reported (Feldman et al., 1994).

Sleep duration

Sleep duration was measured using a body monitoring system (SenseWear Pro2 Armband; BodyMedia, Pittsburgh, Commonwealth of Pennsylvania, USA; http://sensewear.bodymedia.com/SW-Company/SW-Mission-and-Values) during each visit. All of the participants wore the SenseWear armband continuously for 24 hours and then returned it to the same research nurse who performed the assessment at each measurement time point. Because many participants claimed that wearing the armband for more than 24 hours at a time was inconvenient, to avoid case attrition, the participants were asked to wear it for only 24 hours at a time. After sleep duration had been measured, the data were uploaded to a computer. An interrater reliability value of .88 was obtained before data were collected. The findings of previous studies indicate that both short and long sleep durations are predictive of cardiovascular-related diseases (Itani et al., 2017; Jike et al., 2018). In addition, a number of prior studies have defined abnormal sleep duration to include both short and long sleep durations (Gershon et al., 2017; Kuo et al., 2016). In this study, “normal sleep duration” was defined as those with sleep durations of 6.1–8.9 hours per day and “abnormal sleep duration” was defined as those with sleep durations of ≤ 6 hours per day or ≥ 9 hours per day (Gershon et al., 2017; Kuo et al., 2016). The results of prior research support SenseWear as a more reliable method for assessing sleep in patients with obstructive sleep apnea than the current gold standard test (polysomnography; Sharif & Bahammam, 2013).

Covariates

A questionnaire was used to collect demographic information from the participants, with educational level treated as a covariate. Educational level response options were as follows: illiterate (coded as 0), certification of attendance at primary school (coded as 1), and primary school or above (coded as 2). Depression symptoms were measured using the Geriatric Depression Scale–Short Form (Herrmann et al., 1996), which comprises 15 items with yes/no responses. The participants replied to the questions based on their emotional response, with scores of 0–5 defined as no depression (coded as 0) and scores of 6–15 defined as indicative of depression (coded as 1). The reliability (internal consistency) and construct validity of the Geriatric Depression Scale–Short Form have been established, and it is widely used in community settings (Chen et al., 2019). Physical pain was assessed using a numeric rating scale from 0 to 10, with 0 indicating no pain and 10 indicating the worst pain imaginable (Paice & Cohen, 1997). This scale is an easily used and effective instrument for evaluating pain intensity (Kim & Jung, 2020). The participants were asked to rate their pain (postsurgery) by circling the number that best described their average pain during the current month. Depression symptoms and physical pain were measured during the first, third, and sixth months after discharge.

Results Participant Characteristics

Of the 55 participants, most were female (n = 42, 76.4%). The mean age of the sample was 77.71 (SD = 7.76) years, 52.7% had received a diagnosis of femoral neck fracture, and 30.9% had undergone ORIF. The average duration of diabetes in the sample was 13.65 (SD = 12.13) years, and 80% of the participants were using oral medicine to control their diabetes (Table 1).

Table 1 - Demographics of the Hip-Fractured Older Adults With Diabetes Mellitus (N = 55) Variable n % Gender  Male 13 23.6  Female 42 76.4 Age (years; M and SD) 77.71 7.76 Marital status  Married 27 49.1  Widower or widow 27 49.1  Divorced 1 1.8 Educational level  Illiterate 29 52.7  Certification of attendance at primary school 11 20.0  Primary school or above 15 27.3 Comorbidities (M and SD) 1.35 8.21 Chronic histories a  Parkinsonism 6 10.9  Dementia 6 10.9  Hypertension 39 70.9  Stroke 13 23.6 Type of fracture  Intertrochanteric 26 47.3  Femoral neck 29 52.7 Surgical methods  Bipolar hemiarthroplasty 14 25.5  Moore 12 21.8  ORIF 17 30.9  CRIF 12 21.8 Duration of diabetes (M and SD) 13.65 12.13 Diabetes control  Diet only 5 9.1  Oral medicine 44 80.0  Insulin 6 10.9

Note. ORIF = open reduction with internal fixation; CRIF = closed reduction with internal fixation.

a Multiple choices.

In the first month after discharge, the average sleep duration was 8.38 hours per day (range: 2.36–15.29 hours per day, SD = 3.33), and nearly 70% (n = 32, 69.5%) of the participants were affected by abnormal sleep duration. In the third month after discharge, the average sleep duration was 7.41 hours per day (range: 2.36–14.13 hours per day, SD = 2.59), and slightly over half of the participants (n = 16, 51.6%) were affected by abnormal sleep duration. In the sixth month after discharge, the average sleep duration was 6.79 hours per day (range: 2.28–12.18 hours per day, SD = 2.21), and 10 (47.6%) of the participants were affected by abnormal sleep duration (Table 2).

Table 2 - Postoperative Changes in Sleep Duration Variable Time After Discharge 1 Month 3 Months 6 Months n % Mean (Hours/Day) SD n % Mean (Hours/Day) SD n % Mean (Hours/Day) SD Total 46 100.0 8.38 3.33 31 100.0 7.41 2.59 21 100.0 6.79 2.21  Range 2.36–15.29 2.36–14.13 2.28–12.18 Normal sleep duration 14 30.4 15 48.4 11 52.4 Short sleep duration 10 21.8 4 12.9 6 28.6 Long sleep duration 22 47.8 12 38.7 4 19.0
Predictors of Abnormal Sleep Duration

The GEE model achieved convergence after adjusting for time, educational level, depression, and pain. The results showed that older adults with more comorbidities had a higher risk of reporting abnormal sleep duration (OR = 3.14, 95% CI [1.05, 9.37], p = .04). Neither age nor gender was found to predict abnormal sleep duration. Although older adults who had undergone ORIF (OR = 2.65, 95% CI [0.80, 4.50], p = .005) or CRIF (OR = 1.39, 95% CI [0.04, 2.74], p = .04) were found to face a higher risk of abnormal sleep duration than their peers who had undergone bipolar hemiarthroplasty, none of these procedures were found to be predictive of abnormal sleep duration. A longer duration of DM was found to increase the risk of abnormal sleep duration (OR = 1.18, 95% CI [1.03, 1.35], p = .01), whereas the participants with DPN were found to be 9.60 times more likely to experience abnormal sleep duration (OR = 9.60, 95% CI [1.39, 66.37], p = .02) than their peers without DPN. Finally, the participants with diabetes-related peripheral vascular disease were found to be 15.62 times more likely to experience abnormal sleep duration (OR = 15.62, 95% CI [2.24, 108.97], p = .006) than their peers who did not have this disease. Interestingly, methods of DM control were not shown to predict abnormal sleep duration (Table 3).

Table 3 - Adjusted ORs (95% CI) for Having an Abnormal Sleep Duration (N = 55) Predictor Incident Abnormal Sleep Duration
(≤ 6 h or ≥ 9 h) ORs a 95% CI p Time (vs. first month)  3 months 0.24 [0.06, 0.89] .03  6 months 0.12 [0.04, 0.37] < .001 Demographic factors  Age 0.92 [0.82, 1.02] .12  Female gender (vs. male gender) 0.22 [0.04, 1.14] .07  Comorbidities 3.14 [1.05, 9.37] .04 Fracture-related factors  Diagnosis (vs. femoral neck fracture): intertrochanteric fracture 0.004 [0.07, 0.21] .08  Surgical method (vs. bipolar): Moore 1.02 [0.13, 7.73] .98  Surgical method (vs. bipolar): ORIF 2.65 [0.80, 4.50] .005  Surgical method (vs. bipolar): CRIF 1.39 [0.04, 2.74] .04 Diabetes-related factors  Duration of diabetes 1.18 [1.03, 1.35] .01  Method of diabetes control (vs. diet control)   Oral medication 0.22 [0.02, 2.48] .21   Insulin 0.52 [0.02, 13.26] .69  DPN (vs. none) 9.60 [1.39, 66.37] .02  Diabetes-related peripheral vascular disease (vs. none) 15.62 [2.24, 108.97] .006 Covariates  Educational level (vs. illiterate)   Certification of attendance at primary school 0.54 [0.11, 2.69] .45   Primary school or above 0.003 [0.001, 0.09] .001  Depression 1.70 [0.37, 7.75] .49  Pain 0.83 [0.54, 1.27] .39

Note. ORs = odds ratios; CI = confidence interval; h = hours; ORIF = open reduction with internal fixation; CRIF = closed reduction with internal fixation; DPN = diabetic peripheral neuropathy.

a Regression coefficients were obtained after controlling for time, educational level, depression, and pain.

Discussion

This is the first longitudinal study conducted to assess the predictors of abnormal sleep duration in older adults with hip fracture and DM. The results support that participants with more comorbidities face a higher risk of abnormal sleep duration, which echoes the findings of a previous systematic review study indicating individuals with physical illness as more likely to experience abnormal sleep (Smagula, Stone, et al., 2016). The results also showed that those participants who had undergone internal fixation (ORIF or CRIF) were more likely to have abnormal sleep duration than those who had undergone bipolar hemiarthroplasty. Although prior studies have not reported a correlation in patients with hip fracture between different types of surgery and hours of sleep, a systematic review study revealed that different types of surgical treatments had different impacts on quality of life in patients with hip fracture (Wantonoro et al., 2020). In addition, a prior study reported that patients with hip fracture who had undergone hemiarthroplasty seemed to recover better in terms of earlier ambulation and less pain than those who had undergone ORIF (Park et al., 2015). In other words, patients with hip fracture who had undergone internal fixation (ORIF or CRIF) seem to have poorer postoperative recovery. ON the basis of Park et al.'s (2015) study findings, older adults with hip fracture and DM who had undergone ORIF or CRIF were hypothesized in this study to be more affected by adversely influenced sleep and thus be more at risk of abnormal sleep duration.

In this study, individuals with a longer duration of DM and those having DPN or diabetes-related peripheral vascular disease were found to face a higher risk of abnormal sleep duration. However, the method used to control DM was not found to predict sleep duration. The findings of several prior studies support a negative impact of DM complications on sleep conditions, including increased rates of obstructive sleep apnea, decreased sleep efficiency, and more sleep fragmentation (Bahnasy et al., 2018; Meng et al., 2016; Smagula, Koh, et al., 2016). This may explain why the participants in this study with a longer duration of DM or with DM complications were identified as more likely to experience abnormal sleep duration their peers. In this study, the lack of correlation found between method of DM control and abnormal sleep duration differs from prior studies. For example, Barakat and colleagues found that patients who used insulin therapy were more likely to have abnormal sleep than their peers who did not use insulin therapy (Barakat et al., 2019). On the other hand, Bahnasy et al. found no correlation between diabetes treatment regimens and sleep conditions (Bahnasy et al., 2018). These inconsistencies in the literature may reflect variations in the methodologies used (e.g., target populations, sleep measurement methods).

Limitations

Several limitations affect this study and its findings. First, the sample size was quite small. Second, only the 24-hour sleep duration at each measuring point was assessed, and subjective sleep quality was not measured. Thus, sleep conditions may not have been fully captured. Third, blood sugar level data were not collected, which disallowed consideration of its impact on sleep duration in the target population. Finally, the predictors for short and long sleep hours were not considered independently because of small numbers in some predictors. This approach thus would not have captured potential differences in predictors between long and short sleep durations.

We suggest that future studies include more participants, use both objective and subjective measures of sleep pattern to detect sleep changes over time, and further explore the correlation between blood sugar levels and sleep duration in older adults with hip fracture who had undergone surgery. We also suggest that future studies of this issue explore predictors for short and long sleep duration separately.

Conclusions

The results of this study indicate that older adults with hip fracture and DM who have more comorbidities, receive internal fixation, or have a long history of DM, DPN, or diabetes-related peripheral vascular disease face a higher risk of abnormal sleep duration than those without these conditions. Several clinical implications may be made based on these findings. First, sleep duration in older adults with hip fracture and a long history of DM and related complications who have undergone internal fixation should be given greater attention to promote better postoperative recovery. Second, interventions should be developed and prescribed to reduce the symptoms of DPN and diabetes-related peripheral vascular disease to promote normal sleep duration in these patients.

Acknowledgments

This study was supported by the National Health Research Institutes, Taiwan (NHRI-EX103-9905PI) and the Chang Gung Medical Foundation, Taiwan (BMRP297). The parent study has been registered at ClinicalTrials.gov (trial number: NCT01051830).

Author Contributions

Study con