Factors associated with injurious from falls in people with early stage\n Parkinson's disease

Sarini Abdullah, James McGree, Nicole White, Kerrie Mengersen, Graham KerrarXiv 2019 · DOI · arXiv

Abstract

Falls are common in people with Parkinson's disease (PD) and have detrimental effects which can lower the quality of life. While studies have been conducted to learn about falling in general, factors distinguishing injurious from non-injurious falls are less clear. We develop a two-stage Bayesian logistic regression model to model the association of falls and injurious falls with data measured on patients. The forward stepwise selection procedure was used to determine which patient measures were associated with falls and injurious falls, and Bayesian model averaging (BMA) was used to account for uncertainty in this variable selection procedure. Data on 99 patients for a 12-month time period were considered in this analysis. Fifty-five percent of the patients experienced at least one fall, with a total of 335 falls cases; 25% of which were injurious falls. Fearful, Tinetti gait, and previous falls were the risk factors for fall/non-fall, with 77% accuracy, 76% sensitivity, and 76% specificity. Fall time, body mass index, anxiety, balance, gait, and gender were the risk factors associated with injurious falls. Thus, attaining normal body mass index, improving balance and gait could be seen as preventive efforts for injurious falls. There was no significant difference in the risk of falls between males and females, yet if falls occurred, females were more likely to get injured than males.

Introduction

Falls are a major clinical concern in Parkinson’s disease (PD), often leading to a significant decline in quality of life. While existing research has extensively documented the risk factors for falling in general, there is a lack of clarity regarding what differentiates a "simple" fall from an injurious one. Most studies treat falls as a binary event (faller vs. non-faller) based on a single occurrence. However, PD patients often experience repeated falls. This paper addresses the need for a more granular, longitudinal analysis that accounts for repeated measures and the inherent uncertainty in selecting which clinical variables actually drive injury risk.

Methods

The authors developed a two-stage Bayesian logistic regression model to distinguish between the risk of falling and the risk of sustaining an injury:

  1. Stage 1: Modeled the probability of a patient being a "faller" vs. a "non-faller."
  2. Stage 2: For those who fell, modeled the probability of the fall being injurious vs. non-injurious. This stage included random effects to account for patient-specific tendencies across multiple fall events.

To handle the complexity of variable selection, they employed Forward Stepwise Selection, Bayesian Model Averaging (BMA) to account for model uncertainty, and Integrated Nested Laplace Approximation (INLA) for efficient computational approximation of posterior distributions.

Datasets and Injury Characteristics

The study analyzed data from 99 patients diagnosed with idiopathic PD (early stage, Hoehn and Yahr score ≤ 3) over a 12-month prospective period using monthly falls diaries. Clinical metrics included the Tinetti (balance and gait), Timed Up and Go (TUG), and psychological metrics like the Beck Anxiety Inventory (BAI).

A total of 335 falls were recorded, with 25% classified as injurious. The injuries ranged from common bruises to more severe outcomes like broken wrists and lacerations.

Table 1: Frequency and types of injuries recorded during the 12-month study period.

Key Insights

  • Predictors of Falling vs. Injury: The factors that make you fall aren't necessarily the ones that get you hurt. Stage 1 identified fearfulness, poor gait (Tinetti), and previous fall history as primary drivers of falling. Stage 2 found that BMI, anxiety, gender, and fall timing were more relevant to injury.
  • The "Afternoon Danger Zone": Falls occurring between 12 p.m. and 9 p.m. had a significantly higher rate of injury (34%) compared to morning falls (16%). This may be due to increased activity levels or waning medication effects.
  • The Gender Gap: Interestingly, there was no significant difference in the risk of falling between males and females. However, when a fall did occur, females were significantly more likely to be injured.
  • The Self-Rating Paradox: Patients who rated their balance as "excellent" or "good" sometimes showed a higher likelihood of injurious falls, suggesting a potential "overconfidence" effect in early-stage PD.

Results

  • Model Performance: The Stage 1 model achieved 77% accuracy, while the Stage 2 model achieved 78% accuracy. Incorporating BMA slightly improved the Area Under the Curve (AUC) for both stages (0.82 and 0.83, respectively).
  • Preventive Strategies: Higher BMI and lower gait scores were strongly associated with injurious falls, suggesting that weight management and physical therapy targeting gait are viable preventive strategies.

Limitations and Future Work

Key limitations include the relatively small cohort size and reliance on general health surveys (SF-36) rather than PD-specific tools. Future work could integrate wearable sensors to provide objective data on fall mechanics and explicitly track medication "on/off" cycles at the time of each fall.

Summary updated

Abstract

Falls are common in people with Parkinson's disease (PD) and have detrimental\neffects which can lower the quality of life. While studies have been conducted\nto learn about falling in general, factors distinguishing injurious from\nnon-injurious falls are less clear. We develop a two-stage Bayesian logistic\nregression model was used to model the association of falls and injurious falls\nwith data measured on patients. The forward stepwise selection procedure was\nused to determine which patient measures were associated with falls and\ninjurious falls, and Bayesian model averaging (BMA) was used to account for\nuncertainty in this variable selection procedure. Data on 99 patients for a\n12-month time period were considered in this analysis. Fifty five percent of\nthe patients experienced at least one fall, with a total of 335 falls cases;\n25% of which were injurious falls. Fearful, Tinetti gait, and previous falls\nwere the risk factors for fall/non-fall, with 77% accuracy, 76% sensitivity,\nand 76% specificity. Fall time, body mass index, anxiety, balance, gait, and\ngender were the risk factors associated with injurious falls. Thus, attaining\nnormal body mass index, improving balance and gait could be seen as preventive\nefforts for injurious falls. There was no significant difference in the risk of\nfalls between males and females, yet if falls occurred, females were more\nlikely to get injured than males.\n
BibTeX
@misc{abdullah2019factors,
  author = {Sarini Abdullah and James McGree and Nicole White and Kerrie Mengersen and Graham Kerr},
  title = {{Factors associated with injurious from falls in people with early stage\textbackslash\{\}n Parkinson's disease}},
  year = {2019},
  eprint = {1910.02379},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1910.02379},
}
Figure