Preventing falls and bone fractures in hospital care is an important issue in geriatric medicine. Use of hypnotics is a potential risk factor for falls and bone fractures in older patients. However, data are lacking on the association between use of hypnotics and the occurrence of bone fracture.
We used a national inpatient database including 1,057 hospitals in Japan and included dementia patients aged 50 years or older who were hospitalized during a period of 12 months between April 2012 and March 2013. The primary outcome was the occurrence of bone fracture during hospitalization. Use of hypnotics was compared between patients with and without bone fracture in this matched case-control study.
Of 140,494 patients, 830 patients suffered from in-hospital fracture. A 1:4 matching with age, sex and hospital created 817 cases with fracture and 3,158 matched patients without fracture. With adjustment for the Charlson comorbidity index, emergent admission, activities of daily living, and scores for level walking, a higher occurrence of fractures were seen with short-acting benzodiazepine hypnotics (odds ratio, 1.43; 95% confidence interval, 1.19–1.73; P<0.001), ultrashort-acting non-benzodiazepine hypnotics (1.66; 1.37–2.01; P<0.001), hydroxyzine (1.45; 1.15–1.82, P=0.001), risperidone and perospirone (1.37; 1.08–1.73; P=0.010). Other drug groups were not significantly associated with the occurrence of in-hospital fracture.
Citation: Tamiya H, Yasunaga H, Matusi H, Fushimi K, Ogawa S, Akishita M (2015) Hypnotics and the Occurrence of Bone Fractures in Hospitalized Dementia Patients: A Matched Case-Control Study Using a National Inpatient Database. PLoS ONE 10(6): e0129366. https://doi.org/10.1371/journal.pone.0129366
Academic Editor: Martijn van Griensven, Klinikum rechts der Isar - Technical University Munich - TUM, GERMANY
Received: February 10, 2015; Accepted: May 7, 2015; Published: June 10, 2015
Copyright: © 2015 Tamiya et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Data Availability: All relevant data are within the paper.
Funding: The authors have no support or funding to report.
Competing interests: The authors have read the journal's policy and have the following competing interests: MA received research grant support by Daiichi-Sankyo and Takeda, and honoraria for educational lectures by Daiichi-Sankyo Takeda, Eisai, Novartis and Mochida. SO received honoraria for educational lectures by Abbott and Daiichi-Sankyo. Other authors declare no conflict of interest. This does not alter the authors' adherence to PLOS ONE policies on sharing data and materials.
Bone fracture following falls in hospitalized patients is an unresolved problem in geriatric medical management. Older patients, especially dementia patients, have higher risks of delirium, insomnia and day—night reversal with environmental changes related to hospital admission. Hypnotics and psychoactives are sometimes used to relieve these symptoms.
Studies have examined the risk of hypnotic drug use on fall and fracture. In a meta-analysis of 22 studies, fall was significantly associated with use of sedatives and hypnotics, neuroleptics and antipsychotics, antidepressants, and benzodiazepines, but was not associated with narcotics . A retrospective analysis of 3683 patients demonstrated that fall was associated with use of hypnotics . Several studies showed that zolpidem was significantly associated with a higher risk of fall and fracture [3, 4]. Another study showed a hip fracture risk associated with non-benzodiazepine hypnotics . These previous studies were mostly performed in nursing homes or in the community in older patients. Falls in the acute care hospital setting occur at a higher rate than falls in nursing home settings [6–9]. Moreover, hospital falls are also most frequent in safety incident reports, and sometimes lead to negligence suits . However, to our knowledge, there has been no study that simultaneously assessed various types of hypnotics and their risk of in-hospital fracture in a nationwide clinical setting. In this study, we performed a matched case-control study to analyze the association between use of various types of hypnotics and the occurrence of bone fracture in hospitalized dementia patients, using a national inpatient database in Japan. We focused on dementia for several reasons. First, older patients with dementia have a fall risk twice or more than that of older patients without dementia [11, 12]. Second, older patients with dementia also have three to eight times more injuries with falls and more often have a bad prognosis [13–15]. Third, wandering, as well as other behavioral and psychological symptoms of dementia, along with hypnotic and psychoactive agents used to relieve these symptoms, increases fall risks [16, 17]. The medications tend to become the problem in acute care hospitals in trying to prevent falls in dementia patients.
Setting and Participants
For this study, we used the Diagnosis Procedure Combination database. The details of the database have been described elsewhere [18–20]. Briefly, the database includes administrative claims data and discharge abstract data, collected from about 1000 participating hospitals across Japan. The database includes the following information: patient age and sex; main diagnoses, comorbidities at admission and complications after admission recorded according to the International Classification of Diseases, 10th Revision (ICD-10) and text data in Japanese; medical procedures; medicines and devices used; length of stay; activities of daily life (ADL) scores at admission; and discharge status. A unique identifier was used for each hospital. All patient identifiers were removed from this database. Because of the anonymous nature of the data, the Institutional Review Board of The University of Tokyo waived the need for written informed consent from the participants. Study approval was also obtained from the Institutional Review Board of The University of Tokyo.
Among approximately 7 million inpatients over 12 months between April 1, 2012 and March 31 2013, we identified patients aged 50 years or older. Of these, we selected patients who were diagnosed with dementia or dementia-related diseases, including dementia in Alzheimer disease (ICD-10 code, F00), vascular dementia (F01), dementia in Pick disease (F02.0, G31.0), dementia in Creutzfeldt—Jakob disease (F02.1, G810), dementia in Huntington disease (F02.2, G10), dementia in Parkinson disease (F02.3, G20), dementia in human immunodeficiency virus disease (F02.4, B220), dementia in other specified diseases classified elsewhere (F02.8), unspecified dementia (F03), alcoholic dementia (F107), dementia in cerebral lipidosis (F028, E756), Lewy bodies dementia (F028, G318), and mild cognitive disorder (F06.7).
We extracted data on the following 17 types of hypnotics (or sedatives used as hypnotics) for each patient: benzodiazepine anxiolytics; diazepam; ultrashort-acting benzodiazepine hypnotics; short-acting benzodiazepine hypnotics; middle- to long-acting benzodiazepine hypnotics; ultrashort-acting non-benzodiazepine hypnotics; melatonin-receptor agonists; hydroxyzine; phenothiazine antipsychotics; haloperidol; sulpiride; risperidone and perospirone; multi-acting-receptor-targeted antipsychotics used as a hypnotic; antidepressants used as a hypnotic; Japanese kampo herbal medicine used as a hypnotic and in the treatment of behavioral and psychological symptoms of dementia; and other neurological drugs used as hypnotics.
Comorbidities were assessed by ICD-10 codes and converted into scores to calculate the Charlson comorbidity index (CCI) based on Quan’s algorithm . ADL scores for walking on a flat floor were also extracted including bedridden (Score 0), totally assisted (Score 1), partially assisted (Score 2) and without disability (Score 3).
The outcome in this study was in-hospital fracture. In the database, comorbidities already present at admission are clearly differentiated from complications that occurred after admission. In-hospital fracture was defined as fracture that occurred after admission and was determined according to the following ICD-10 codes: fracture of skull and facial bones (S02); fracture of neck (S12); fracture of rib(s), sternum and thoracic spine (S22); fracture of lumbar spine and pelvis (S32); fracture of shoulder and upper arm (S42); fracture of forearm (S52); fracture at wrist and hand level (S62); fracture of femur (S72); fracture of lower leg, including ankle (S82); fracture of foot, except ankle (S92); fractures involving multiple body regions (T02); fracture of spine, level unspecified (T08); fracture of upper limb, level unspecified (T10); and fracture of lower limb, level unspecified (T12).
We performed a matched case—control study. First, we identified cases with in-hospital fracture. For each case, we selected four controls of similar age (±5 years) and the same sex from the same hospital. When there were more than four matched-control candidates to each case, we randomly selected four control patients. Specifically, control cases were sorted by randomly generated values from a Microsoft SQL server and the top four were selected. There are two ways to conduct matching: matching with replacement; and matching without replacement . Matching with replacement means that controls can be used as matches for more than one treated individual; matching without replacement signifies that controls cannot be used as matches for more than one treated individual. Though the statistical analysis becomes more complex, matching with replacement can often decrease bias because controls that resemble many treated individuals can be used multiple times [23, 24]. Moreover, the order of matching the treated individuals is immaterial in the case of matching with replacement. One methodological paper compared matching with and without replacement in three matching methods, and it found that matching with replacement had a smaller bias among all three methods . Thus, we chose matching with replacement for the present study. If a control case was a candidate for more than one case, we included both matches. In the following analysis, one control was selected three times, 95 controls were selected twice, and they were weighted using frequency weights. If the number of matched-control candidates for each case was less than four, we also included both the corresponding case (62 cases) and control (138 controls) in the analytical group subset to avoid selection bias, unless no control subjects were assigned (13 cases).
Descriptive statistics were presented for the matched patients. Categorical variables were compared using the chi square test. We performed multivariable logistic regression for the occurrence of in-hospital fractures fitted with a generalized estimating equation to account for the clustered nature of the cases and controls. There are two ways to cluster in matched case-control studies: generalized estimating equations (GEEs) and conditional logistic analysis. Both methods can make consistent estimates. As GEE is more robust in terms of the specification of matching effect, we chose GEEs . The dependent variable was in-hospital fracture, and independent variables included, emergent admission, ADL score for walking on a flat floor, CCI and 17 classes of drugs. All statistical analyses were conducted using IBM SPSS version 22.0 (IBM SPSS, Armonk, NY, USA).
Among 140,494 eligible patients, 830 patients suffered from in-hospital fracture.
Using 1:4 matching, we obtained a case group of 817 patients and a control group of 3158 patients. Table 1 shows the baseline characteristics of the matched patients (n = 3975). As a result of matching, there was no significant difference in age (P = 0.582) or sex (P = 0.728) between the case and control groups. To exclude the possibility that controls may have been matched to cases with a larger age difference than to cases with a smaller age difference, we also compared the distribution of age in the case and control groups. Mean, median, standard deviation, range and interquartile range of age (years) in the case and control groups were 81.5 vs. 81.8, 82.0 vs. 82.0, 7.9 vs. 7.5, 50–103 vs. 50–103, 11.0 vs. 10.0, respectively. The age distributions of the case and control groups were also similar. No significant difference in CCI or emergent admission was present between the cases and controls. The ADL score for walking on a flat floor on admission was significantly different between the groups.
Table 2 shows 17 types of hypnotics (or sedatives used as hypnotics) used for the case and the control groups. The most frequently used drugs in both groups were ultrashort-acting non-benzodiazepine hypnotics, followed by short-acting benzodiazepine hypnotics. The proportion of patients who used any of the 17 types of drugs was significantly higher in the case than the control group (66.8% vs. 51.9%, P<0.001) (Table 2). The proportion of patients who used more than three types of hypnotics (or sedatives used as hypnotics) was also higher in the case than the control group (24.2% vs. 14.6%, P<0.001). The case group was significantly more likely to use benzodiazepine anxiolytics; ultrashort-acting benzodiazepine hypnotics; short-acting benzodiazepine hypnotics; middle- to long-acting benzodiazepine hypnotics; ultrashort-acting non-benzodiazepine hypnotics; melatonin-receptor agonists; hydroxyzine; phenothiazine antipsychotic; haloperidol; sulpiride; risperidone and perospirone; multi-acting-receptor-targeted antipsychotics used as a hypnotic; and an antidepressant used as a hypnotic. The proportion of patients who used Japanese kampo herbal medicine was not significantly different between the cases and controls.
Table 3 shows the results of the multivariable logistic regression analysis. A higher occurrence of in-hospital fracture was significantly associated with use of a short-acting benzodiazepine hypnotic, an ultrashort-acting non-benzodiazepine hypnotic, a hydroxyzine, risperidone and perospirone. Neither melatonin agonists nor Japanese kampo herbal medicine was associated with the occurrence of in-hospital fracture.
The present study showed an increased risk for in-hospital fracture with several hypnotics and psychoactives used as hypnotics in dementia patients who were admitted to acute care hospitals. In-hospital fracture risk was associated with a short-acting benzodiazepine hypnotic, an ultrashort-acting non-benzodiazepine hypnotic, and risperidone and perospirone.
A previous US study of nursing home residents demonstrated a fracture risk with ultrashort-acting non-benzodiazepine hypnotics . The present study also showed a risk with ultrashort-acting non-benzodiazepine hypnotics. A new finding in the present study is that shorter-acting drugs had relatively higher odds ratios for in-hospital fracture than longer-acting drugs. A possible explanation for this may be that patients are more likely to fall when drowsy soon after taking hypnotics. Another possible reason is that physicians may have avoided prescribing long-acting hypnotics for frail patients.
To our knowledge, the present study is the first to show that hydroxyzine may increase the risk of fracture in dementia patients. Hydroxyzine has been shown to be as effective as bromazepam, one of the benzodiazepines, in the treatment of generalized anxiety disorder . Because hydroxyzine has a half-life of around 3 hours, it may act like a short-acting benzodiazepine.
Our results showed that a melatonin agonist was not significantly associated with the occurrence of in-hospital fracture. Melatonin-receptor agonists including ramelteon are new types of hypnotics. They act on the GABAA receptor-independent pathway in contrast to most of hypnotics that act on the GABAA receptor. The present study suggests that a melatonin agonist may be safer than other hypnotics in terms of fall and fracture risk.
Japanese kampo herbal medicines, which are often used as sedatives or hypnotics in Japan, have beneficial effects on the behavioral and psychological symptoms of dementia [28–30]. Our results suggest that these drugs would be good alternatives to conventional hypnotics or sedatives in dementia patients and may reduce fracture risk.
This study has several limitations. First, recorded diagnoses in an administrative claims database are less well validated than those in planned prospective studies. Second, the time interval between drug administration and related in-hospital fracture cannot be identified from the database and its causal relationship remains to be clarified. We have information on the timing and use of these agents, but not on the timing of fracture. Consequently, we are not certain as to whether the agent was prescribed before or after the fracture, or if it was not used for a short period of time for weeks to months prior to fracture. Third, it is difficult to distinguish the deleterious effect of hypnotic use itself from underlying conditions, including night delirium and insomnia requiring prescription of hypnotics. Fourth, there was no information about previous falls, and so we were unable to examine this relationship owing to the lack of data.
In light of these findings, it is preferable to avoid prescribing short-acting benzodiazepines and ultrashort-acting non-benzodiazepine hypnotics, risperidone or perospirone, hydroxyzine, or multi-acting-receptor-targeted antipsychotics to in-hospital dementia patients. Melatonin-receptor agonists or Japanese kampo herbal medicine may be preferable to these drugs.
Short-acting benzodiazepines and ultrashort-acting non-benzodiazepine hypnotics were associated with an increase in in-hospital fractures in dementia patients, while no significant association with an increase in in-hospital fractures was seen with middle- to long-acting benzodiazepine hypnotics, melatonin-receptor agonists, or Japanese kampo herbal medicine.
Conceived and designed the experiments: HT HY SO. Performed the experiments: HT HY SO. Analyzed the data: HT HY SO. Contributed reagents/materials/analysis tools: HM KF HY. Wrote the paper: HT HY MA SO.
- 1. Woolcott JC, Richardson KJ, Wiens MO, Patel B, Marin J, Khan KM, et al. Meta-analysis of the impact of 9 medication classes on falls in elderly persons. Archives of internal medicine. 2009;169(21):1952–60. Epub 2009/11/26. pmid:19933955.
- 2. Obayashi K, Araki T, Nakamura K, Kurabayashi M, Nojima Y, Hara K, et al. Risk of falling and hypnotic drugs: retrospective study of inpatients. Drugs in R&D. 2013;13(2):159–64. Epub 2013/06/14. pmid:23760758; PubMed Central PMCID: PMC3689908.
- 3. Kolla BP, Lovely JK, Mansukhani MP, Morgenthaler TI. Zolpidem is independently associated with increased risk of inpatient falls. Journal of hospital medicine: an official publication of the Society of Hospital Medicine. 2013;8(1):1–6. Epub 2012/11/21. pmid:23165956.
- 4. Wang PS, Bohn RL, Glynn RJ, Mogun H, Avorn J. Zolpidem use and hip fractures in older people. Journal of the American Geriatrics Society. 2001;49(12):1685–90. Epub 2002/02/15. pmid:11844004.
- 5. Berry SD, Lee Y, Cai S, Dore DD. Nonbenzodiazepine sleep medication use and hip fractures in nursing home residents. JAMA internal medicine. 2013;173(9):754–61. Epub 2013/03/06. pmid:23460413; PubMed Central PMCID: PMC3676706.
- 6. Halfon P, Eggli Y, Van Melle G, Vagnair A. Risk of falls for hospitalized patients: a predictive model based on routinely available data. Journal of clinical epidemiology. 2001;54(12):1258–66. Epub 2001/12/26. pmid:11750195.
- 7. Tutuarima JA, van der Meulen JH, de Haan RJ, van Straten A, Limburg M. Risk factors for falls of hospitalized stroke patients. Stroke; a journal of cerebral circulation. 1997;28(2):297–301. Epub 1997/02/01. pmid:9040678.
- 8. Rubenstein LZ, Josephson KR, Robbins AS. Falls in the nursing home. Annals of internal medicine. 1994;121(6):442–51. Epub 1994/09/15. pmid:8053619.
- 9. Morgan VR, Mathison JH, Rice JC, Clemmer DI. Hospital falls: a persistent problem. American journal of public health. 1985;75(7):775–7. Epub 1985/07/01. pmid:4003655; PubMed Central PMCID: PMC1646291.
- 10. Oliver D, Killick S, Even T, Willmott M. Do falls and falls-injuries in hospital indicate negligent care—and how big is the risk? A retrospective analysis of the NHS Litigation Authority Database of clinical negligence claims, resulting from falls in hospitals in England 1995 to 2006. Quality & safety in health care. 2008;17(6):431–6. Epub 2008/12/10. pmid:19064658.
- 11. Tinetti ME, Speechley M, Ginter SF. Risk-Factors for Falls among Elderly Persons Living in the Community. New Engl J Med. 1988;319(26):1701–7. pmid:ISI:A1988R516200004.
- 12. van Doorn C, Gruber-Baldini AL, Zimmerman S, Hebel JR, Port CL, Baumgarten M, et al. Dementia as a risk factor for falls and fall injuries among nursing home residents. Journal of the American Geriatrics Society. 2003;51(9):1213–8. pmid:ISI:000185321900003.
- 13. Oleske DM, Wilson RS, Bernard BA, Evans DA, Terman EW. Epidemiology of injury in people with Alzheimer's disease. Journal of the American Geriatrics Society. 1995;43(7):741–6. Epub 1995/07/01. pmid:7602023.
- 14. van Dijk PT, Meulenberg OG, van de Sande HJ, Habbema JD. Falls in dementia patients. The Gerontologist. 1993;33(2):200–4. Epub 1993/04/01. pmid:8468012.
- 15. Magaziner J, Simonsick EM, Kashner TM, Hebel JR, Kenzora JE. Predictors of functional recovery one year following hospital discharge for hip fracture: a prospective study. Journal of gerontology. 1990;45(3):M101–7. Epub 1990/05/01. pmid:2335719.
- 16. Buchner DM, Larson EB. Falls and fractures in patients with Alzheimer-type dementia. Jama. 1987;257(11):1492–5. Epub 1987/03/20. pmid:3820464.
- 17. Asada T, Kariya T, Kinoshita T, Asaka A, Morikawa S, Yoshioka M, et al. Predictors of fall-related injuries among community-dwelling elderly people with dementia. Age and ageing. 1996;25(1):22–8. Epub 1996/01/01. pmid:8670525.
- 18. Sumitani M, Uchida K, Yasunaga H, Horiguchi H, Kusakabe Y, Matsuda S, et al. Prevalence of malignant hyperthermia and relationship with anesthetics in Japan: data from the diagnosis procedure combination database. Anesthesiology. 2011;114(1):84–90. Epub 2010/12/21. pmid:21169796.
- 19. Hamada T, Yasunaga H, Nakai Y, Isayama H, Horiguchi H, Matsuda S, et al. Continuous regional arterial infusion for acute pancreatitis: a propensity score analysis using a nationwide administrative database. Crit Care. 2013;17(5):R214. Epub 2013/10/04. pmid:24088324; PubMed Central PMCID: PMC4055985.
- 20. Iwagami M, Yasunaga H, Doi K, Horiguchi H, Fushimi K, Matsubara T, et al. Postoperative polymyxin B hemoperfusion and mortality in patients with abdominal septic shock: a propensity-matched analysis. Critical care medicine. 2014;42(5):1187–93. Epub 2013/12/25. pmid:24365858.
- 21. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. Journal of chronic diseases. 1987;40(5):373–83. Epub 1987/01/01. pmid:3558716.
- 22. Austin PC. A critical appraisal of propensity-score matching in the medical literature between 1996 and 2003. Statistics in medicine. 2008;27(12):2037–49. Epub 2007/11/27. pmid:18038446.
- 23. Stuart EA. Matching methods for causal inference: A review and a look forward. Statistical science: a review journal of the Institute of Mathematical Statistics. 2010;25(1):1–21. Epub 2010/09/28. pmid:20871802; PubMed Central PMCID: PMC2943670.
- 24. Dehejia RH, Wahba S. Causal effects in, nonexperimental studies: Reevaluating the evaluation of training programs. J Am Stat Assoc. 1999;94(448):1053–62. pmid:ISI:000083879800007.
- 25. Zhao Z. Using matching to estimate treatment effects: Data requirements, matching metrics, and Monte Carlo evidence. Rev Econ Stat. 2004;86(1):91–107. pmid:ISI:000220322400007.
- 26. Lin IF, Lai MY, Chuang PH. Analysis of matched case-control data with incomplete strata: applying longitudinal approaches. Epidemiology. 2007;18(4):446–52. Epub 2007/05/26. pmid:17525695.
- 27. Llorca PM, Spadone C, Sol O, Danniau A, Bougerol T, Corruble E, et al. Efficacy and safety of hydroxyzine in the treatment of generalized anxiety disorder: a 3-month double-blind study. The Journal of clinical psychiatry. 2002;63(11):1020–7. Epub 2002/11/26. pmid:12444816.
- 28. Iwasaki K, Satoh-Nakagawa T, Maruyama M, Monma Y, Nemoto M, Tomita N, et al. A randomized, observer-blind, controlled trial of the traditional Chinese medicine Yi-Gan San for improvement of behavioral and psychological symptoms and activities of daily living in dementia patients. The Journal of clinical psychiatry. 2005;66(2):248–52. Epub 2005/02/12. pmid:15705012.
- 29. Mizukami K, Asada T, Kinoshita T, Tanaka K, Sonohara K, Nakai R, et al. A randomized cross-over study of a traditional Japanese medicine (kampo), yokukansan, in the treatment of the behavioural and psychological symptoms of dementia. Int J Neuropsychopharmacol. 2009;12(2):191–9. Epub 2008/12/17. pmid:19079814.
- 30. Okahara K, Ishida Y, Hayashi Y, Inoue T, Tsuruta K, Takeuchi K, et al. Effects of Yokukansan on behavioral and psychological symptoms of dementia in regular treatment for Alzheimer's disease. Progress in neuro-psychopharmacology & biological psychiatry. 2010;34(3):532–6. Epub 2010/02/23. pmid:20170698.