回復期リハビリテーション病棟の退院時FIMは入院時FIMでどの程度予測可能か?—重回帰分析と交差妥当性による検討

How Predictive is the FIM at Discharge from Convalescent Rehabilitation Ward by the FIM at Admission? : A Study Using Multiple Regression Analysis and Cross-validation

要旨　目的：本研究の目的は，回復期リハビリテーション病棟に入院するすべての患者に適用可能で，入院初期に評価可能な単純な指標を説明変数とするADLの予測モデルを構築することである．

方法：対象は当院の回復期リハビリテーション病棟に入院した1,153名の患者であった．入院時のFIMの18の下位項目，性別，年齢，発症からの期間を説明変数とし，退院時の13の運動下位項目を合計して算出したFIM運動スコア合計を目的変数として，ステップワイズ重回帰分析を行った．予測モデルが臨床的に適用可能であることを確認するため，解析対象とは独立した85名の被験者に対し交差妥当性の検証を行った．

結果：解析群の重回帰分析の結果，R²が0.712と有意な重回帰式が得られた．また，検証群の85名の患者について，退院時の予測FIMと実際のFIMの差を両群間で検定したところ，両群間に有意差はなく，r=0.888（p＜0.01）の強い相関が認められた．

結論：この予測モデルは，目的変数が簡便に取得でき予測式を構築する労力が少ない．本研究の結果から，他の回復期ハビリテーション病棟においても各々のデータを用いて予後予測モデルを構築し，臨床応用することに資する可能性が示唆された．

Abstract　Purpose：We aimed to build a predictive model for ADL, applicable to all patients admitted to a convalescent rehabilitation ward, that uses a simple index as explanatory variables, which is accessible early during hospitalization.

Methods：We included 1153 patients admitted to our convalescent rehabilitation ward. Stepwise multiple regression analysis was conducted using 18 Functional Independence Measure (FIM) sub-items at admission, with sex, age, and time since onset as explanatory variables. The total FIM motor score was calculated by summing 13 motor sub-items at discharge. Cross-validation was performed on 85 participants, independent of the analysis group, to confirm the model's clinical applicability.

Results：The multiple regression analysis results for the analysis group showed a significant equation with an R2 of 0.712. In the validation group with 85 patients, difference between the predicted and actual FIM scores at discharge was compared between the two groups, and a strong correlation of r=0.888 (p＜0.01) was observed, without any significant difference between the groups.

Conclusion：This model is less labor intensive when constructing a prediction equation because the objective variable is easily obtained. Our findings indicate that other convalescent rehabilitation wards can develop prognostic models using their data and implement them clinically.