Risk of accidental injuries amongst Parkinson disease patients.

Risk of accidental injuries amongst Parkinson disease patients

H.-C. Wang ^a,b,c , C.-C. Lin ^d , C.-I. Lau ^a , A. Chang ^a , F.-C. Sung ^d,e and C.-H.

Kao ^e,f

a

Department of Neurology, Shin Kong Wu Ho-Su Memorial Hospital, Taipei;

College of Medicine, National Taiwan University, Taipei;

c

College of Medicine, Taipei Medical University, Taipei;

Management Office for Health Data, China Medical University Hospital,

Taichung;

Graduate Institute of Clinical Medicine Science and School of Medicine, College of Medicine, China Medical University,

Taichung; and

Department of Nuclear Medicine and PET Center, China Medical University Hospital, Taichung, Taiwan

Introduction

Parkinson disease (PD) is a common neurodegenerative disease characterized by bradykinesia, rigidity and resting tremors. In its advanced stages, non-L-dopa responsive problems, including falls and associated fractures, become major problems for PD patients [1].

Fall rates are extremely high amongst PD patients.

The findings of several studies have revealed that 38%–87% of parkinsonian patients experienced falls [1–3]. Almost 70% of patients fell during a 1-year follow- up period. Recurrent falls occurred in approximately 50% of patients during 1 year [4]. Accidental injuries (AIs) are expected amongst PD patients.

Previous studies have shown that PD patients have a high risk of developing fractures, particularly in their hips [5–10]. In addition to fractures, numerous other injuries can result from falls. Aside from falls, other mechanisms such as motor vehicle accidents might also be responsible for some AIs in PD. The risk of injury from motor vehicle accidents was found to be significantly increased for patients with PD [11,12].

The full spectrum and associated risk of injuries experienced

by PD patients has not been evaluated on a

nationwide scale. Thus, the risk of all types of AIs amongst Taiwanese PD patients was investigated.

Methods

Data source

The study cohort was constructed from the Longitudinal

Health Insurance Database (LHID), a subset of the National Health Insurance Research Database

(NHIRD) in Taiwan. The NHIRD contains annual claim data for reimbursement from the Taiwan

National Health Insurance program, which is a universal and single-payer health insurance that covered

99% of the 23 million Taiwan citizens in 2011. The National Health Research Institute maintains and releases data for research. The LHID used in this study contains data of 1 million randomly selected insured subjects between 1996 and 2000. Data from these subjects have then been continuously collected and renewed for research purposes. Based on National Health Research Institute reports, there is no difference in age and sex distribution between LHID and

NHIRD. The details of the LHID 2000 have been examined in previous studies [13].

Ethics statement

Individual identification was encoded to protect patient confidentiality and all investigators signed an agreement guaranteeing this before using the database.

This study was approved by the Institutional

Review Board of the China Medical University Hospital (CMU-REC-101-012).

Study cohorts

In this study, the case group comprised PD patients

newly diagnosed between 2000 and 2009 (ICD-9-CM

332) obtained from the LHID 2000. The index date

for the PD patients was the date of their first medical

visit for PD. Patients who were below 50 years

of age, lacked age and sex information, or had had

AIs within 1 year before the index date were

excluded. Regarding the comparison group, 4-fold

non-PD insurants from the same database of the same period using the same exclusion criteria were selected, frequency-matching them to the case cohort in terms of age (within 5 years) and sex. Hypertension (ICD-9-CM 401–405), diabetes mellitus (ICD-9-

CM 250), coronary artery disease (CAD) (ICD-9- CM 410–414), stroke (ICD-9-CM 430–438), epilepsy (ICD-9-CM 345), injury history before the index date, and dementia were also considered as baseline comorbidities.

End of follow-up in the study cohort

Accidental injuries in the NHIRD include the following injuries: spine and spinal cord (ICD-9-CM 767.4), brachial plexus (ICD-9-CM 767.6), vault of skull or facial bone fracture (ICD-9-CM 800–804.99), vertebra fracture (ICD-9-CM 805–806.9), thoracic cage fracture (ICD-9-CM 807–807.6), pelvis fracture (ICD-9-CM 808–808.9), limb fractures (ICD-9-CM 810–828.1), dislocations (ICD-9-CM 831–839.79), head injuries (ICD-9-CM 850–854.19, and 959.01), superficial injuries and contusions (ICD-9-CM 910–924.9), burns

(ICD-9-CM 940–949.5) and spinal cord and nerve injuries (ICD-9-CM 952–957.9).

The AIs were subsequently stratified into head injuries (ICD-9-CM 800–804, 850–854 and 959.01), bone fractures and dislocations (ICD-9-CM 805–808, 810–828 and 831–839), burns (ICD-9-CM 940–949), injuries to the spinal cord, plexus and nerves (ICD- 9-CM 767.4, 767.6 and 952–957) and superficial injuries and contusions (ICD-9-CM 910–924). The duration of follow-up for each study subject in 100

person-years (100 PYs) was calculated from the baseline date to the date of the AI diagnosis, withdrawal

from the insurance program or 31 December 2010;

whichever occurred first indicated the end of followup.

The AIs were treated as the ‘events’ for the

incidence and survival analyses, and loss of followup

or reaching 31 December 2010 was treated as

censored.

Statistical analysis

Regarding the demographic factors and comorbidities, the mean and standard deviations of the continuous variables were calculated and the number

and proportion of the category variables. To assess the differences between the cohorts, a t test was applied for the continuous variables and a chisquared test for the category variables. The crude

incidence rate was calculated as the number of incident cases divided by the combined PYs from each

individual in the cohort. The cumulative incidence curves were estimated using the Kaplan–Meier method, and the differences between these curves were tested using the log rank test. The risk of AIs in the PD compared with the control group is presented as hazard ratios (HRs) and 95% confidence

intervals (CIs) estimated in a multivariate Cox proportional hazard model adjusted for age, sex, hypertension,

diabetes, CAD, stroke, epilepsy, injury

history and dementia. Incident rate calculations and Cox models were also calculated separately for each level of the factors presented in Table 1. The multivariate models were calculated using SAS software,

version 9.1 (SAS Institute Inc., Cary, NC, USA).

The cumulative incidence curves were produced using the SAS LIFETEST procedure, and then curves were plotted using R. P values <0.05 or CIs for HRs that excluded the value 1.00 were considered statistically significant. Results

Baseline cohort characteristics

During the follow-up period, the PD (4046 patients)

and comparison (16 184 participants) cohorts were

surveyed. The baseline characteristics of these cohorts

were similar with regard to age and sex due to the

selection of a frequency-matched comparison cohort

(Table 1); however, the comorbidities demonstrated

significant differences between the PD and comparison

cohorts regarding hypertension, diabetes, CAD, stroke, epilepsy, injury history and dementia. Patients who lacked comorbidities also demonstrated a significant difference between the PD and comparison

cohorts (20% vs. 9%).

The hazard ratio of subsequent injury for accidental injuries

The PD cohort exhibited a higher incidence rate of subsequent injury than did the comparison cohort regarding head injury (2.95 per 100 PYs vs. 1.42 per 100 PYs), bone fracture and dislocation (4.61 per 100 PYs vs. 3.19 per 100 PYs), burns (0.66 per 100 PYs vs. 0.58 per 100 PYs), injuries to the spinal cord, plexus and nerves (0.15 per 100 PYs vs. 0.12 per 100 PYs) and superficial injuries and contusions

(11.41 per 100 PYs vs. 9.09 per 100 PYs) (Table 2).

These injuries were evaluated after adjusting for age, sex, hypertension, diabetes, CAD, stroke, epilepsy, injury history and dementia. The PD cohort attained a higher overall injury incidence rate (19.78 per 100 PYs vs. 14.4 per 100 PYs) than did the comparison cohort, with a 1.30-fold increased risk of overall AIs (HR 1.30, 95% CI 1.24–1.36) after adjusting the covariates. Figure 1 shows that the PD cohort

attained a significantly higher cumulative incidence of subsequent injury than did the comparison cohort.

As for specific types of AI, the PD cohort was associated with increased risk of head injuries (HR 1.88,

95% CI 1.64–2.15), bone fracture and dislocations (HR 1.39, 95% CI 1.25–1.54) and superficial injury and contusions (HR 1.20, 95% CI 1.12–1.27) compared with the comparison cohort. On the other hand, the risks of burns (HR 1.01, 95% CI 0.78–1.32) or injury to spinal cord, plexus and nerves (HR 1.25, 95% CI 0.72–2.17) was not significantly increased in PD patients compared with the controls.

The hazard ratio of subsequent overall injuries based

on age, sex and comorbidity

When the data were stratified for ages 50–69 years, 69–79 years and 80+ years, the incidence rates

increased in the two older age categories for both cohorts; and the incidence rate of the PD cohort was

higher compared with the comparison cohort for each of the three age groups (Table 3). The adjusted HR of subsequent injury was 1.16–1.36 amongst the three age groups.

After dividing by sex, the PD cohort demonstrated higher incidence rates than did the comparison cohort and the adjusted HR was approximately 1.28–1.32- fold. The patients who lacked comorbidities attained a higher adjusted HR (approximately 1.37) compared with the comorbid patients (95% CI 1.17–1.60). The presence or absence of hypertension, diabetes, CAD and stroke comorbidities yielded similar adjusted HRs (approximately 1.27–1.31) in the PD and comparison cohorts. However, the adjusted HR was 0.99 (95%

CI 0.74–1.33) in epilepsy patients and 1.31 (95%

CI 1.24–1.37) in non-epilepsy patients. The adjusted HR in patients who demonstrated a history of injury was 1.20 (95% CI 1.12–1.29), whereas patients who demonstrated no history of injury attained an HR of 1.39 (95% CI 1.30–1.49). The adjusted HR in patients

who had dementia was 1.23 (95% CI 1.04–1.44), whereas patients who had no dementia attained an

HR of 1.30 (95% CI 1.24–1.37). PD patients have significantly increased risk of AIs compared with controls

under every stratified status, the only exception being epilepsy (Table 3).

Discussion

This is the first study to investigate the spectrum of AIs in PD patients on a national scale by using the NHIRD. Our findings demonstrate that, after adjusting for comorbidities, PD patients in Taiwan demonstrated a higher incidence of subsequent injuries (HR

1.30) compared with the comparison cohort. An elevated

risk level was demonstrated in each injury category

except burn injuries and injuries to spinal cord,

plexus and nerves. Head injuries exhibited the highest HR (1.88), followed by bone fractures and dislocations (1.39) and superficial injuries and contusions

(1.20). The HR of spinal cord, plexus and nerve injuries was 1.25. However, its 95% CI includes zero and thus does not definitively indicate increased risk. To our knowledge, the risk of head injury following PD diagnosis has not been reported; however, several studies have implicated head injury as a possible PD etiology. Numerous studies have demonstrated an increased prevalence of head injury before the onset of PD [14–17]. A recent study concluded that the increased frequency of head injury during the months preceding PD diagnosis is a consequence of the evolving movement disorder rather than its cause [18]. Our findings suggest that PD patients demonstrate early head injuries because the PD cohort comprised newly diagnosed patients. The results agree with those of previous studies, in which falls occurred in the earliest stages of PD [2]. PD patients have shown multiple prodromal symptoms and decline of quality of life

before diagnosis [19,20]; thus it is possible that PDrelated head injuries attained before the onset of PD

reflect prodromal functional deficits. Our results show that PD patients have the highest risk of developing head injuries, which warrant the most attention of all AIs. Head injury has a substantially higher adjusted HR (1.88-fold) than does bone fracture and dislocation (1.39-fold).

As previous studies have suggested, the risk of bone fracture and dislocation is also substantial.

The risk of superficial injuries and contusions are also elevated amongst PD patients, attaining an adjusted HR of 1.20-fold. By contrast, the risk of burns is not increased in PD patients (1.01 HR).

The most common causes of burn injuries are scalds

involving hot oil or water, thermal burns resulting

from fires or kitchen gas explosions, and electrical burns [21–23]. Although it is presumed that PDrelated falls had contributed to a large portion of

AIs, other common mechanisms of AIs may also have roles in the increased AI risks in PD. For

example, the risk of injury from motor vehicle accidents was found to be significantly increased for

patients with PD [11,12]. Because of the lack of associated information in the NHIRD database, it was not possible to establish the contribution of motor vehicle accidents as well as that of other potential causes including violence and accidents involving drugs and alcohol.

Distinct PD patient groups are at risk of various injuries. The adjusted HR of the 69–79 age group of PD patients relative to age-matched controls (1.38) was significantly higher than was that of the 50–69 age groups (1.16), and the HR slightly decreased in

the ≥80 age groups (1.33). This is consistent with previous studies, which have reported that fall risk

increases as age increases, noting the complex Ushaped relationship, i.e. a paradoxical decrease of fall

risk amongst the most advanced cases, between falls and disease severity [24,25]. These results suggest that clinicians should consider how to prevent falls and subsequent immobilization in PD patients of advanced age.

All comorbidities (hypertension, diabetes, CAD,

stroke, epilepsy, injury history and dementia) were significantly higher in the PD cohort than in the comparison

cohort. By contrast, the PD cohort comprised

significantly fewer participants who lacked comorbidities (9.1% vs. 19.9%) than did the comparison cohort.

It is suspected that this indicates an intrinsic problem regarding NHIRD-based studies. The PD cohort primarily comprised chronic patients who regularly visited

clinics and hospitals; thus, their comorbid

illnesses were recorded in the database. The comparison

cohort inevitably included numerous patients who exhibited minor ailments or visited medical facilities less; thus, their comorbidities were not recorded. This suggests that NHIRD-based studies must involve adjusting hazards relative to baseline comorbidities.

How much impact this discrepancy of comorbidity has on our results is not known. However, if there had been a way to acquire more accurate information to better correct for comorbidities in the comparison group and thus identify a higher proportion of controls with comorbidities, the differences of hazards for AIs between the PD and comparison cohorts might have been even larger after adjustment. After exploring the adjusted injury HRs regarding specific comorbidities, a consistently elevated injury risk amongst

the PD cohort was determined, regardless of hypertension,

diabetes, CAD or stroke. Notably, PD patients who demonstrated a history of injuries showed a

lower relative injury risk (1.20 HR) than did patients

who lacked a history of injuries (1.39 HR). This phenomenon may reflect the fact that the control patients

with previous injuries were already at higher risk of

subsequent injury, and thus the high risk in the PD

cohort no longer produced a high relative risk relative

to the control group. On the other hand, epileptic PD

patients exhibited a decreased adjusted HR (0.99) in

contrast with non-epilepsy patients (1.31). In this case,

because the actual rate of injuries was lower (18.72%)

in epileptic PD patients compared with non-epilepsy

patients (19.83%) (Table 3), it indicates a true

decrease of relative injury risk in epileptic PD

patients. It is speculated that epileptic PD patients

would have decreased mobility. Decreased walking

and ambulation may lead to decreased chances of

accidents and can explain this seemingly paradoxical

phenomenon. Cognitive impairment is another issue

that should be addressed. There is strong evidence

that global measures of cognition are associated with

serious injury [26]. Theoretically cognitive impairment as a comorbidity would further increase the risk of AIs in PD patients. However, our results show a slightly decreased adjusted HR (1.23 vs. 1.30) as well as a lower actual rate of injuries (19.62% vs. 19.81%) in dementia PD patients compared with non-dementia patients. As in the case of epilepsy comorbidity, this seemingly paradoxical protective effect of dementia may only reflect decreased mobility and more dependence in dementia PD patients.

This study is subject to certain limitations that must be mentioned. First, the NHIRD does not provide

detailed patient information regarding alcohol consumption, body mass index, physical activity levels

and socioeconomic status, all of which are possible confounding factors. Secondly, evidence derived from a cohort study is typically subject to numerous biases related to adjusting confounders. A key limitation

remains despite our meticulous study design and adequate control of confounding factors: possible unmeasured or unknown confounders could generate a bias.

Thirdly, our study cohort comprised PD patients at their first medical visits. Those dates do not represent the onset of PD symptoms but could be months or even years after initial symptoms. Finally, the National Health Insurance claim registries primarily function as administrative billing instruments and are not verified for scientific purposes. It was not possible to contact patients directly to obtain additional information because of the anonymity assured by the identification numbers. However, the data that were

obtained regarding PD diagnoses and injury definitions

were highly reliable.