Monday, September 17, 2018

Hip Fracture Decisions for Nursing Home Residents with Dementia

by Bob Arnold (@rabob)

Good things come in threes. I was on service this week and saw a patient with Lewy Body dementia in the emergency room after he fell and broke his hip. His niece was his surrogate and trying to decide what to do. She asked me whether he should have his hip repaired. In looking through the literature I came across Sarah Berry’s article “Association of Clinical Outcomes of Surgical Repair of Hip Fractures versus Nonsurgical Management in Nursing Home Residents with Advanced Dementia”. Even better, there was an editorial in the same issue of JAMA Internal Medicine placing the article in context. Second, as I was reading the article, I heard the trauma surgeon talking about the same article with her residents. Finally, this morning I woke up and read Bree Johnston’s fabulous special report in JPM on hip fractures in the setting of limited life expectancy, “The importance of considering goals of care and prognosis”. Yahtzee! (My children would point out that only a true nerd would find this trio of events a cause for celebration). While I would recommend you read both the editorial and Dr. Johnston’s fabulous special report, the purpose of this Pallimed series is to look at the data so I’ll be reviewing Berry’s article.

The aim of the article was to assess outcomes for nursing home residents with advanced dementia who did and did not undergo surgical repair of a broken hip. Advanced dementia was defined as a Cognitive Performance Scale (CPS) of 5 or 6 and diagnosis of “dementia” or “Alzheimer disease.” By outcomes they meant survival (the primary outcome), pain, anti-psychotic use, physical restraints, pressure ulcers, and ambulatory status. The authors used the MDS (Minimum Data Set) assessment and linked this to Medicare claims to conduct a cohort study of 3,083 long-staying nursing home residents with advanced dementia and hip fracture.

They found, as have other studies of nursing home residents with hip fractures, that patients who underwent surgery had lower mortality rates (the literature stressed the importance of surgery within the first 24 hours). While 35% died within 6 months, and 61% within two years, the mortality was significantly greater in patients who did not have surgery. These results were greatest in the first 30 days; 11% mortality with surgery, 30% without surgery, and resulted in a median survival of 1.4 years with surgery versus 0.4 years if the patient did not undergo surgery. Adjustment attenuated the findings, but they remained significant.

Residents who underwent surgical repair also had less pain, less anti-psychotic drug use, physical restraint use and pressure ulcers; although once these results were adjusted for differences between the two groups there were no differences according to surgical repair. Interestingly the inverse probability of treatment waiting models, which adjusted for differences in characteristics before the hip fractures, suggested that there was less pain and fewer pressure ulcers among patients managed with surgery. (I need someone who knows more about statistics than me to explain why two different ways of statistical correction resulted in different secondary outcomes).

So, the question is how I should use these results in my patient:

1. Are the patients in this population relevant to the ones I care for? Well, as a hospital-based palliative care doctor these are exactly the kinds of patients for which I am consulted.

2. Are the outcomes that the authors measured the correct ones? Well, it seems to me that pain, survival, and restraints are all things that my patients’ families want to know about. Sadly, for these secondary outcomes they could only look at one point in time. Also, to have data on the secondary outcomes, the patient had to live at least six months (it has to do with when MDS data is collected). Thus, for a lot of the patients we do not have these secondary outcomes.

3. Were the two groups similar in characteristics prior to the operation? Sadly, the answer is no. Residents treated non-operatively were much more impaired at baseline. For example, 26% of the non-operative residents were completely dependent in activities of daily living as opposed to only 5% of the surgical residents. Moreover, despite the large number of variables in the MDS, it is likely there were differences between the two groups that were unmeasured that led the surgeons to choose not to do surgery. In addition, it is unclear whether the decision to do surgery was based on patient/family preferences. Thus, one does not know if the reason for the difference in outcomes was based on surgery or whether other variables led both to the decision to have surgery and the outcomes. This is a limitation of not doing a RCT. Finally, the outcomes could be due to a self-fulfilling hypothesis. Given the non-surgical patients’ greater illness/morbidity, there may have been a decision only to focus on comfort. Given this, the treatment these patients received was less focused on prolonging life and thus they died sooner. (This would have nothing to do with the impact of surgery on survival or clinical outcomes).

4. Were the circumstances and methods for detecting the outcome similar? The answer here is yes. The MDS is a very complete way of detecting the outcomes of interest. Although pain was evaluated by the health care provider rather than the patients, there is no reason to think that there would be differences based on which group they were in.

5. Was follow-up sufficiently complete? Again, the answer seems to be yes, although as previously noted, for patients who did not live six months we do not have any of the secondary outcomes.

6. Are the differences big enough that I should care? Again, the answer is at least with the primary outcome the difference did seem quite large as noted before, the secondary outcomes, differences, particularly after adjustment, are much smaller.

So, what does this mean? In the end, as Johnston et al. summarized, the decision to have surgery depends a great deal on the surrogate decision-maker’s view about the patient’s quality of life preoperatively and what is most important postoperatively. I have to say that this article would, for many of my patients, lead me to do surgery and continue aggressive palliative care (the increased rate of ambulation postoperatively -10.7% in the patients with surgery versus 4.8% in those without surgery - would be a big factor for many families). While a randomized controlled file would be better, it is unlikely that one will ever be conducted. (I am given pause by the mostly negative data presented by Johnston on hip fracture repair. She points out, for example, there is the Cochran review of five randomized controlled trials that shows no difference in medical complications, mortality or long-term pain in conservative care versus surgery. While this is not a study of demented patients, it made me realize the data is controversial).

A coda: What I and the editorial found distressing were the high rates of pain and the low rates of hospice in severely demented patients even after they have hip fractures (particularly in the non-operative patients). It also was quite curious that the median time to utilize hospice was 56 days. Given this article, I wonder whether hospice should be discussed and/or recommended for all patients who have severe dementia and a hip fracture. This article should lead you to talk to your trauma surgeons and/or orthopedists to develop a routine palliative care or hospice consultation for these patients.

Robert Arnold, MD is a palliative care doctor at the University of Pittsburgh and a co-founder of VitalTalk (@VitalTalk). He loves both high and low brow comedy (The Good Place and Nanette), pop culture (the National Enquirer and Pop Culture Happy hour) and music of all kinds (not opera tho!) You can find him on Twitter at @rabob. 

More Pallimed posts from Bob Arnold can be found here. More journal article reviews can be found here. 


1. Berry SD, Rothbaum RR, Kiel DP, Lee Y, Mitchell SL. Association of clinical outcomes with surgical repair of hip fracture vs nonsurgical management in nursing home residents with advanced dementia [published online May 7, 2018]. JAMA Intern Med. doi:10.1001/jamainternmed.2018.0743

2. Mehr DR, Tatum PE, Crist BD. Hip Fractures in Patients With Advanced Dementia What Treatment Provides the Best Palliation? JAMA Intern Med. 2018;178(6):780–781. doi:10.1001/jamainternmed.2018.0822

3. Johnston CB, Holleran A, Ong T, McVeigh U, Ames E. Hip Fracture in the Setting of Limited Life Expectancy: The Importance of Considering Goals of Care and Prognosis. Journal of Palliative Medicine 2018 21:8, 1069-1073

4. Morrison RS, Siu AL. Survival in end-stage dementia following acute illness. JAMA. 2000;284(1):47-52

Pallimed | Blogger Template adapted from Mash2 by Bloggermint