台灣地區六十五歲以上老年族群健康狀態影響死亡率的性別差異研究

Abstract

研究緣起及目的 目前對於65歲以上老年人缺乏綜合性、方便性及準確的工具適用於預測老人的健康狀況。我們有興趣於了解性別對平均壽命的影響，及兩性是否有顯著危險因子分布的不同；希望藉由評估老人的社會心理方面、生物醫學因子測量、生理疾病方面及功能方面的不同危險因子，分析男女性別間是否存在重要風險因子差異導致最終的預期壽命的不同。基於以上原因，所以我們希望利用中華民國老年醫學會於1989年7月1日至1991年6月30日間對台灣地進行社區65歲以上健康調查資料，描述台灣地區65歲以上老人的健康狀況及危險因子的分布，並比較性別之間的差異，及研究各個因子對老人健康的影響。 研究材料及方法

我們設計了一個前瞻性的世代研究 (prospective cohort study)，利用此世代在1989-1991間所收集的四個方向的健康資訊包括: 年齡、性別、性行為、性行為頻率、性慾、寡居狀態、巴氏量表評分、巴氏量表個別項目評分、長谷川氏癡呆量表評分、身體質量指數、及生化指標等。將全死因及特別死因的死亡作為結果。所我們連結了2003年的台灣地區的死亡登記計算追蹤時間。我們追蹤此一研究世代直至2003年12月31日，平均追蹤時間為11年。利用生命表法描繪存活曲線，利用Cox regression model估計各變項的死亡風險比。最後將我們收集到的健康資訊，設計一個含上述四個方面危險因子的死亡預測風險模式，將危險因子的參數代入我們發展的模式中，估算老人的存活時間。 結果

兩性之間有統計差異而且盛行率較高或數值較高的風險因子在女性較高的為： 寡居狀態、BMI、收縮壓、總膽固醇、低密度脂蛋白、三酸甘油酯、糖尿病史。在男性為性活力、長谷川氏痴呆量表評分、尿酸、肌酸酐、血紅素、飲酒習慣、吸菸習慣及中風病史。 我們利用存活函數積分計算出的65歲的平均餘命 (life expectancy)分別為全體：16.17年，女性：17.84年，男性：14.68年。 控制了所有的相關危險因子後我們發現有性行為的男性全死亡風險下降了33% (adjusted hazard ratio, aHR=0.67 (95% CI:0.56-0.80))；女性則下降了16% (aHR=0.84 (95% CI:0.65-1.09))；不分性別則整體下降了28% (aHR=0.72 (95% CI:0.62-0.84))。所以性行為是影響老年族群全死因及特別死因的獨立危險因子，而且無性別上的差異。高頻率性行為組及中頻率組的存活率皆優於低頻率組，但在高頻率組及中頻率組中間則無明顯統計差異。 有性慾的男性減少19%了死亡風險 (aHR= 0.81 (95% CI: 0.68~0.97)，但在女性則無此關聯性。 寡居狀態的個案有較高的死亡風險比 (aHR= 1.43 (95% CI: 1.21~1.68))，這種關係在男女性皆存在 (男性：aHR= 1.66 (95% CI: 1.25~2.19)；女性：aHR= 1.33 (95% CI: 1.09~1.62))。 當個案有日常生活功能障礙時，調整干擾因子後其全死因死亡風險比正常個案高出73% ( aHR=1.73 (95%CI: 1.44-2.01)) 。若我們依不同性別作計算，仍然可以得到相似的結果，我們發現有日常生活功能障礙的男性多了70%的死亡風險 (aHR =1.70 (95%CI: 1.32-2.20)) ；女性則多了72% (aHR =1.72 (95%CI: 1.30-2.28)) 。 巴氏量表上任何一個項目異常都會明顯增加其全死因死亡風險，影響最明顯的是上廁所能力障礙 (toilet use) ，其HR=3.31 (95% CI: 2.52~3.72)；影響最小的是小便控制能力障礙 (bladder) ，但其死亡風險比仍然很高 (HR=2.17 (95% CI: 1.77-2.66)) 。若我們調整了所有干擾因子，自我修飾功能障礙 (grooming) 有最高的死亡風險比 (aHR= 2.43 (95% CI: 1.46-4.04)；最不明顯為穿衣能力障礙 (dressing) ，其aHR=1.57 (95% CI: 1.00-2.45) 。 當個案有心智功能障礙時，調整干擾因子後其全死因死亡風險比正常個案高出33% (aHR =1.33 (95%CI: 1.16-1.52)) 。若我們依不同性別作計算，仍然可以得到相似的結果，我們發現有心智功能障礙的男性多了33%的死亡風險 (aHR=1.33 (95%CI: 1.13-1.57)) ；女性也多了33% (aHR=1.33 (95%CI: 1.06-1.69)) 。 BMI對台灣65歲以上老人的全死因死亡風險的影響在女型呈U形分布，男性呈L形分布。體重過輕在不同性別都具有較高的死亡率。體重過重在兩性都有保護作用，肥胖在女性是危險因子但在男性反而有保護作用。吸煙不會干擾這種關係。另外BMI對死亡率的影響主要在短期，隨追蹤時間愈長，其影響愈不明顯。 血清白蛋白愈高則對老人的健康具有保護的作用。血清白蛋濃度最高組、中高組、中低組對最低組的死亡調整年齡後的全死因死亡風險比分別為：0.86 (95% CI: 0.81-0.92)，0.80 (95% CI: 0.74-0.87)，0.70 (95% CI:0.60-0.80) 。可見較高的血清白蛋白對健康具有保護作用。在特別死因的分析上，男性並無特殊的差異；但在女性白蛋白最低組死於糖尿病的相對危險性是其他三組的2.02倍 (RR:2.02, 95%CI: 1.12-3.68), 另外死於肝癌的相對危險性是4.10倍(RR: 4.10, 95% CI: 1.59-10.48)。 血清尿酸濃度中高組、中低組、最低組等三組與最高組相比之間的死亡age adjusted hazard ratio，分別為1.14 (95% CI:1.08-1.20)， 1.15 (95% CI:1.06-1.25)，1.24 (95% CI:1.06-1.46) ，高的尿酸值有較高的死亡率。但尿酸值有性別上的差異，男性的尿酸值在最高及最低的兩個quartiles有較高的死亡率；女性則只有最高的quartiles有較高的死亡率。在死因方面的分析，無論男女性尿酸最高組死於中風的相對危險性高於其他三組 (男性RR:1.72, 95%CI:1.10-2.71；女性 RR=1,91, 95% CI: 1.12-3.27) 。另外女性尿酸最高組有較高的相對風險死於腎臟病 (RR=2.91, 95% CI: 1.27-6.67) 及子宮頸癌 (RR: 4.07, 95%CI: 1.08-15.3)。男性尿酸最低組的死因和其他三組並無差別。 我們參考上述研究結果後，在我們的風險預測模式中放入包含四個方面危險因子的年齡、性別、性行為、寡居狀態、日常功能性障礙、心智障礙、身體質量指數、白蛋白、血紅素、尿酸、肌酸酐、麩氨酸-焦葡萄酸轉胺酶 (Glutamic pyruvic transaminase,GPT)、三酸甘油酯、吸菸習慣、冠狀動脈心臟病史、糖尿病史、中風病史等十七個變項，估計老人的存活時間。 結論 男女性別間的危險因子分布有相當大的不同，但這些危險因子絕大部分對兩性的影響力是類似的；只有在BMI及血清尿酸的作用在兩性並不相同。 女性的風險因子種類普遍少於男性，僅有無性行為、低智能量表評分、較高的平均BMI、高三酸甘油酯，高糖尿病病史及低血紅素等，這是女性較健康的原因。但我們在調整很多慢性病的變因後女性仍然有較低的死亡率，可見性別本身對健康就是一個非常重要的風險因子。 雖然我們已能利用我們的研究結果發展出風險預測模式，但仍需要再加入一些未知的有相關的變項才能更準確地預測個案的未來的死亡機率。

Introduction and aim of research There are few tools for predicting health condition of people older than 65 years old now, and we have interesting in how gender factor affects lifespan and distribution of risk factors of mortality in elderly people. We hope to realize the relationship by analyzing risk factors about psychosocial aspect, biologic measurement, physical aspect, and functional aspect. Materials and methods

We designed a prospective cohort study. A total of 2600 subjects enrolled from a nationwide survey on health status of residents age 65 years or older in Taiwan between 1989 and 1991 were followed up until 31 December 2003 for ascertaining cause of death. The average time of follow-up year was 11 years. The health information gathered between 1989 to 1991 included age, gender, sexual activity, frequency of sexual activity, libido, widowhood status, Barthel index score, Hasegawa dementia scale, body mass index, anthropometric measures, life-style factors, preexisting disease, biochemical markers. We computed cumulative survival curve to show the effect of risk factors on all-cause death by life table estimates. Adjusted hazard ratios either by age or by all possible confounding factors were also calculated by Cox proportional hazards regression model. At least, we designed a mortality prediction model including the four aspects of risk factors to predict survival time of the elderly people. Result The risk factors that women had higher prevalence or value included widowhood status, body mass index, systolic blood pressure, total cholesterol, low density lipoprotein, triglyceride, diabetes mellitus history; the risk factors that men had higher prevalence and value included sexual activity, Hasegawa dementia scale, serum uric acid, creatinine, hemoglobin, alcohol drinking habit, smoking habit, and stroke history. The life expectancy calculated by survival function was 16.17 years for all people, 17.84 years for women, and 14.68 years for men. After controlling for age and relevant confounding factors, sexual activity may reduce mortality by 33% (adjusted hazard ratio (aHR) =0.67 (95%CI: 0.56-0.80)) for men, 16% (aHR=0.84 (95% CI: 0.65-1.09)) for women, and 28% (aHR=0.72 (95% CI: 0.62-0.84)) for both sex combined. Males reported to have libido may lead to a 19% decrease of mortality (aHR=0.81 (95% CI: 0.68-0.97)). The elevated risk for mortality attributed to widowhood status was estimated as 66% (aHR=1.66 (95% CI: 1.25-2.19)) for men, 33% (aHR=1.33 (95% CI: 1.09-1.62)) for women, and 43% (aHR=1.43 (95% CI: 1.21-1.68)) for both sex combined. Sexual activity has been found to reduce 36% (aHR=0.64 (95% CI: 0.41-1.00)) mortality from stroke. After controlling for age and relevant confounding factors, people with impaired activity of daily life (ADL) function may increase all-cause mortality by 70% (aHR =1.70 (95%CI: 1.32-2.20)) for men, 72% (aHR =1.72 (95%CI: 1.30-2.28)) for women, and 73% ( aHR =1.73 (95%CI: 1.44-2.01)) for both sex combined. Impairment in every item of the Bathel index score will also induce high all-caused mortality. Among all items, we found people with grooming function impairment had the highest adjusted hazard ratio of all-cause mortality (aHR= 2.43 (95% CI: 1.46-4.04)), and had the lowest adjusted hazard ratio (aHR=1.57 (95% CI: 1.00-2.45)) with dressing function impairment. People with cognitive function impairment may increase all-cause mortality by 33% (aHR=1.33 (95%CI: 1.13-1.57)) in men, 33% (aHR=1.33 (95%CI: 1.06-1.69)) in women, and 33% (aHR =1.33 (95%CI: 1.16-1.52)) for both sex combined. The relationship between body mass index and all-cause mortality was L-shape in men, but U-shape in women. Underweight men had always highest mortality than other BMI groups, but the relationship was seen only in the third and fifth years in women. Overweight women had the lowest mortality at all stages, but the lowest mortality occurred in overweight or obesity men. Obesity women showed higher mortality than normal range group. We found BMI had no effect on mortality of elderly people if we re-analyzed after removing people died in the first five years. The effect of BMI on mortality of elderly people would reduce gradually when the time of follow-up prolonged. The relationship between body mass index and all-cause mortality on non-smoker is U-shape in women, but changed by time in men. Underweight men still had the highest mortality after the fifth years of follow-up. Higher serum albumin level had protective effect on elderly people. The age adjusted hazard ratio of all-cause mortality of the first, second, and third quartiles of serum albumin level contrast to the lowest quartile were 0.86 (95% CI: 0.81-0.92), 0.80 (95% CI: 0.74-0.87), and 0.70 (95% CI:0.60-0.80) . The relative risk of women died from diabetes mellitus in the lowest quartile contrast to others was 2.02 (95%CI: 1.12-3.68), and the relative risk of women died from hepatocellular carcinoma was 4.10 (95% CI: 1.59-10.48). Higher serum uric acid level had higher all-cause mortality on elderly people. The age adjusted hazard ratio of all-cause mortality of the first, second, and third quartiles of serum uric acid level contrast to the lowest quartile were 1.14 (95% CI:1.08-1.20), 1.15 (95% CI:1.06-1.25), 1.24 (95% CI:1.06-1.46) for both sex combined. The similar effect was seen in women, but the effect of serum uric acid level on all-cause mortality was U-shape in men. The highest quartile and the lowest quartile had higher mortality than other two quartiles. People in the highest quartile of serum uric acid level had higher mortality rate died form stroke than others (men: RR: 1.72, 95%CI: 1.10-2.71 ; women: RR=1,91, 95% CI: 1.12-3.27) . Otherwise, women in the highest quartile had higher risk died from kidney disease (RR=2.91, 95% CI: 1.27-6.67) and cervical cancer (RR: 4.07, 95%CI: 1.08-15.3). According to our studying result, we designed an all-cause mortality prediction model, and the parameters of model included age, gender, sexual behavior, widowhood status, ADL function, cognitive function, body mass index, serum albumin, hemoglobin, serum uric acid, creatinine, GPT (glutamic pyruvic transaminase), triglyceride, smoking habit, coronary heart disease history, diabetes mellitus history, and stoke history. Conclusion Distribution of risk factors of mortality between men and women was different, but the effects of these risk factors on mortality were similar in both gender, except body mass index and serum uric acid level. Except without sexual activity, low Hasegawa dementia scale, high average body mass index, high triglyceride level, high prevalence of diabetes mellitus, and low average hemoglobin, women had fewer risk factors than men. This can explain the reason why women had lower mortality rate than men. After adjusting all confounders, we still found women had lower mortality than men, so female gender is a very important health protection factor. Although we had developed a mortality prediction model for elderly people, we still need find more unknown risk factors to improve the ability of predicting mortality of our model.