Dementia is associated with increased mortality and poor patient-centered outcomes after vascular surgery

Abstract

Objective:

Dementia has been associated with increased complications and mortality in orthopedics and other surgical specialties, but has received limited attention in vascular surgery. Therefore, we evaluated the association of dementia with surgical outcomes for elderly patients with Medicare who underwent a variety of open and percutaneous vascular surgery procedures.

Methods:

We reviewed claims data from the Centers for Medicare and Medicaid Services for beneficiaries enrolled in Medicare Part A fee-for-service insurance from January 1, 2011, to December 31, 2011, who underwent inpatient vascular surgery. Only the first surgery during the first admission was considered for analysis. Traditional outcomes (30- and 90-day mortality, intensive care admission, complications, length of stay) and patient-centered outcomes (discharge to home, extended skilled nursing facility [SNF] stay, time at home) were adjusted for patient and procedure characteristics using multilevel linear or logistic regression as appropriate. All analyses were performed using SAS (v9.4, SAS Institute Inc, Cary, NC).

Results:

Our study included 210,918 patients undergoing vascular surgery, of whom 27,920 carried a diagnosis of dementia. The average age of the entire cohort was 75.74 years, and 55.43% were male. Patients with dementia were older and had higher rates of comorbidities compared with patients without a dementia diagnosis. The three most common defined classes of intervention excluding miscellaneous ones were cerebrovascular, peripheral arterial, and aortic cases, which jointly accounted for 53.15% of cases. Among all cases, 56.62% were open. Emergent/urgent cases were more frequent amongst those with dementia (60.66% vs 37.93%; P < .001). After adjustment, patients with dementia had increased odds of 30-day mortality (odds ratio [OR], 1.21; P < .0001) and 90-day mortality (OR, 1.63; P < .0001), extended SNF stay (OR, 3.47; P < .0001), and longer hospital length of stay (8.29 days vs 5.41 days; P < .001). They were less likely to be discharged home (OR, 0.31; P < .0001) and spent a lower fraction of time at home after discharge (63.29% vs 86.91%; P < .001). Intensive care admission and inpatient complications were similar between the two groups.

Conclusions:

Dementia is associated with poor traditional outcomes, including increased 30- and 90-day mortality and longer hospital lengths of stay in this large national patient sample. It is also associated with worse patient-centered outcomes, including substantially lower discharge rates to home, less time spent at home after discharge, and higher rates of extended stay in a SNF. These data should be used to counsel patients facing vascular surgery to provide goal-concordant care, particularly to patients with dementia.

Dementia encompasses a broad set of conditions characterized by persistent cognitive deficits in memory, reasoning, and other domains that are severe enough to affect normal functioning. The worldwide prevalence of dementia in 2010 was 35.6 million, which is predicted to increase rapidly to 65.7 million by 2030, making it a serious public health concern.¹ Dementia is particularly relevant to the vascular surgery population; common risk factors for both vascular disease and dementia include age, smoking, diabetes, and obesity.² Indeed, Partridge et al³ found that 68% of patients undergoing vascular surgery at a single institution had cognitive impairment, which included dementia and lesser forms of disease. Furthermore, this was previously unrecognized in 88% of patients in this latter group.³

Beyond a shared epidemiology, some studies have begun to delineate the negative impact on outcomes after surgery in patients with dementia. Bai et al,⁴ for instance, demonstrated increased mortality after surgery for hip fracture in patients with dementia, and others have shown increased rates of pneumonia, surgical site infection, and mortality after a variety of procedures.^5,6 Despite the relevance of dementia to outcomes for vascular surgery patients, research assessing this relationship has been limited. We sought to evaluate the association of dementia with traditional and patient-centered outcomes after a broad variety of open and endovascular procedures using the Medicare fee-for-service data.

METHODS

Data sources and study population.

This study was approved by the Partners institutional review board; individual patient consent was waived. Claims data from the Medicare Provider Analysis and Review file from January 1, 2011, to December 31, 2011, were used for this study. We included all fee-for-service beneficiaries who underwent an inpatient vascular surgery; only the first admission was used for analyses. Vascular surgeries were identified using ICD-9 procedure codes (Appendix, online only). All vascular surgery procedures were included in the study except for nonspecific codes (eg, 38.29, Other diagnostic procedures on blood vessels) and those for lymphatic procedures to yield a final list of 120 procedures. Surgeries were then dichotomized as having either an endovascular or open surgical approach and categorized into 1 of 10 classes, based principally on anatomic location: cerebrovascular, aortic, venous, dialysis related, peripheral arterial, major amputation (ie, at and above the ankle joint), minor amputation (all remaining more distal amputations), nonaortic intrathoracic, nonaortic intra-abdominal, and miscellaneous. In an attempt to create categories with homogeneous procedural intensity, five codes involving veins were reclassified. For instance, 38.39 (resection of vessel with anastomosis, lower limb veins) was classified as peripheral arterial rather than venous, which tends to include low-intensity ambulatory-type procedures. Procedures with a 30-day mortality of at least 1% for the overall group were considered high risk, consistent with prior studies.⁷ Emergent/urgent status included patients with an admission type coded as emergent or urgent in the claims data, excluding transfers from other hospitals or emergency room as the admission source, and excluding an admission type of trauma or newborn. Dementia diagnoses were identified by International Classification of Disease, Ninth Revision (ICD-9) diagnosis codes (ICD-9-CM codes 331.0, 331.1, 331.11, 331.19, 331.2, 331.7, 290.0, 290.10, 290.11, 290.12, 290.13, 290.20, 290.21, 290.3, 290.40, 290.41, 290.42, 290.43, 294.0, 294.1, 294.10, 294.11, 294.8, and 797) used in the Centers for Medicare and Medicaid Services chronic conditions data warehouse. These codes included a broad array of Alzheimer’s disease and related disorders and non-Alzheimer’s dementia and had to be present in at least one inpatient, outpatient, or skilled nursing facility (SNF) encounter.

Statistical analysis.

The risk-adjusted outcomes of the study included both traditional postoperative outcomes (30- and 90-day mortality, intensive care unit (ICU) admission, inpatient complications, length of stay [LOS]), and patient-centered outcomes (discharge to home, extended stay at SNF, home time, discharge to higher level of care). Inpatient complications were defined using ICD-9 codes that have been previously described.⁸ An extended SNF stay consisted of a LOS of at least 100 days at SNF from the admission date to the date of death, the date of discharge, or the end of 2011, whichever came first. Only patients discharged from the hospital from January 1, 2011, to September 22, 2011 (ie, 100 days before the end of the year) were included in this latter calculation. Home time was defined, for patients who were community-dwelling before admission, as the proportion of accumulated days at home from the hospital admission date for the index procedure to the date of censoring or completion of the calendar year. Discharge to a higher level of care was determined using a hierarchy of discharge destinations, defined as follows in descending order: long-term acute care, SNF, and home.

Categorical factors were described using frequencies and percentages. Continuous measures were described using means and standard deviations. Pearson χ² test and robust t-test were performed to determine the unadjusted differences as appropriate. In addition, multilevel multivariable logistic regression analysis was then applied for all dichotomous outcomes to adjust for differences in patient factors (age, sex, race/ethnicity, medical comorbidities) and surgical characteristics (procedure types, approaches and classes). Hospital clustering was accounted for using a random effects model. A similar approach was taken for LOS and home times using multilevel multivariable linear regression. A P value of .05 was used as a threshold for significance. All analyses were performed using SAS (version 9.4; SAS Institute Inc, Cary, NC).

RESULTS

A total of 210,918 patients undergoing vascular surgery were identified, including 27,920 (12.94%) with dementia. The average age of the entire cohort was 75.74 years, and 55.43% were men. There were several differences in demographics between the patients with and without dementia (Table I). Patients with dementia were older with a mean age of 78.75 years compared with 75.28 years in the group without dementia (P < .001). There were also more women in the group with dementia (51.50% vs 43.67%; P < .001). Additionally, the dementia group had a higher proportion of non-white races, 26.94% vs 16.61% (P < .001). A large majority were noninstitutionalized: 93.23%, 1.14%, and 5.51% of patients without dementia came from home, a SNF or intermediate care facility, and miscellaneous sources, respectively. Analogous percentages for those with dementia were 84.42%, 7.44%, and 7.91%. Patients in the dementia group tended to have higher Charlson comorbidity scores (P < .001). Individual comorbidities including diabetes, renal insufficiency, congestive heart failure, myocardial infarction history, chronic obstructive pulmonary disease, stroke history, hepatic insufficiency, and malignancy, including metastatic solid tumors, were all more frequent in those with dementia (all P < .001).

Table I.

Patient and procedure characteristics

	Total (N = 210,918)		Without dementia (n = 182,998)		With dementia (n = 27,920)		P value
Patient characteristics
Mean age, years	75.74		75.28		78.75		<.001
Male sex	116,908	55.43%	103,087	56.33%	13,821	49.50%	<.001
Race							<.001
Non-Hispanic white	175,895	83.39%	155,496	84.97%	20,399	73.06%
Non-Hispanic black	24,851	11.78%	18,948	10.35%	5903	21.14%
Hispanic	3797	1.80%	2980	1.63%	817	2.93%
Asian/Native American	3344	1.59%	2878	1.57%	466	1.67%
Other	2443	1.16%	2164	1.18%	279	1.00%
Unknown	588	0.28%	532	0.29%	56	0.20%
Comorbidities
Diabetes	80,164	38.01%	67,948	37.13%	12,216	43.75%	<.001
Renal Insufficiency	60,723	28.79%	49,934	27.29%	10,789	38.64%	<.001
COPD	53,798	25.51%	46,955	25.66%	6843	24.51%	<.001
History of MI	27,919	13.24%	24,460	13.37%	3459	12.39%	<.001
Congestive heart failure	39,337	18.65%	31,706	17.33%	7631	27.33%	<.001
Procedure characteristics
Approach							.023
Open	119,415	56.62%	103,430	56.52%	15,985	57.25%
Endovascular	91,503	44.38%	79,568	43.48%	11,935	42.75%
Class							<.001
Cerebrovascular	55,967	26.53%	52,300	28.58%	3667	13.13%
Miscellaneous	48,140	22.82%	41,168	22.50%	6972	24.97%
Peripheral arterial	29,705	14.08%	25,609	13.99%	4096	14.67%
Aortic	26,456	12.54%	24,596	13.44%	1860	6.66%
Minor amputation	12,715	6.03%	10,089	5.51%	2626	9.41%
Major amputation	12,240	5.80%	7867	4.30%	4373	15.66%
Diagnostic	11,022	5.23%	8880	4.85%	2142	7.67%
Dialysis related	10,383	4.92%	8607	4.70%	1776	6.36%
Intra-abdominal	2432	1.15%	2168	1.18%	264	0.95%
Intrathoracic	1614	0.77%	1494	0.82%	120	0.43%
Venous	244	0.12%	220	0.12%	24	0.09%
Status
Emergent/urgent	86,348	40.94%	69,412	37.93%	16,936	60.66%	<.001
High risk	54,663	25.92%	49,732	27.18%	4931	17.66%	<.001

COPD, Chronic obstructive pulmonary disease; MI, myocardial infarction.

There were procedural differences between the two groups as well (Table I). Open procedures were slightly more common in the dementia group: 57.25% vs 56.52% (P = .023). The distributions of procedural class were significantly different (P < .001). Notable differences include more minor and major amputations, peripheral arterial, dialysis, and diagnostic procedures and fewer aortic, nonaortic intra-abdominal, and nonaortic intrathoracic procedures (P < .001). Emergent/urgent procedures were significantly more common in those with dementia (60.66% vs 37.93%; P < .001). High-risk procedures were less common in the those with dementia: 17.66% vs 27.18% (P < .0001).

Both traditional and patient-centered outcomes were generally poorer in the dementia group (Table II). After adjustment, patients with dementia had increased 30-day (odds ratio [OR], 1.21; 95% confidence interval [CI], 1.15–1.28; P < .0001) and 90-day mortality (OR, 1.63; 95% CI, 1.57–1.70; P < .0001), hospital LOS (+0.92 days; 95% CI, 0.83–0.99 days; P < .0001), and risk of extended stay (≥100 days) in an SNF (OR, 3.47; 95% CI, 3.23–3.71; P < .0001) (Table II). They spent a lower fraction of time at home after discharge (−14.77%; 95% CI, −15.17% to −14.36%; P < .0001). Discharge to a higher level of care was much more likely for patients with dementia (OR, 3.06; 95% CI, 2.96–3.17; P < .0001). ICU admission and inpatient complications were similar between the two groups. They were less likely to be discharged home (OR, 0.31; 95% CI, 0.30–0.32; P < .0001) (Table III).

Table II.

Unadjusted and adjusted outcomes amongst patients undergoing inpatient vascular surgery

	Unadjusted			Adjusted (ORs or beta coefficient, 95% CI)
	Without dementia	With dementia	P value	Without dementia	With dementia	P value
Traditional postoperative outcomes
30-Day mortality	5.62%	10.29%	<.001	REF	1.21 (1.15–1.28)	<.0001
90-Day mortality	9.46%	21.24%	<.001	REF	1.63 (1.57–1.70)	<.0001
Inpatient complications	11.44%	14.12%	<.001	REF	1.04 (1.00–1.09)	.0686
ICU admission	48.00%	54.09%	<.001	REF	1.03 (0.99–1.08)	.13
Hospital LOS, days	5.41	8.29	<.001	REF	+0.92 (0.83–0.99)	<.0001
Patient-centered postoperative outcomes
Discharge to home	75.58%	38.26%	<.001	REF	0.31 (0.30–0.32)	<.0001
Discharge to higher level of care	12.77%	42.07%	<.001	REF	3.06 (2.96–3.17)	<.0001
Extended SNF stay	1.93%	10.74%	<.001	REF	3.47 (3.23–3.71)	<.0001
Home-time ratio	86.91%	63.29%	<.001	REF	−14.77 (−15.17 to −14.36)	<.0001

CI, Confidence interval; ICU, intensive care unit; LOS, length of stay; OR, odds ratio; SNF, skilled nursing facility.

Table III.

Discharge destination^a

Destination	Without dementia, %	With dementia, %
Home	75.58	38.26
SNF/LTAC	19.67	54.43
Death	3.22	3.85
Hospice	0.96	2.42
Other	0.56	1.04

SNF/LTAC, Skilled nursing facility/long-term acute care.

^aOverall comparison between patients with and without dementia, unadjusted P < .001.

DISCUSSION

High-quality health care must be concordant with patient values and preferences. Communication of outcomes data is a logically necessary element for such care. Decision making in patients with dementia in particular may be challenging for numerous reasons, including an inability of the patient to fully participate, the presence of multiple formal and informal surrogates, and the lack of a consensus approach on how best to approach surgical decision making with these patients. Therefore, in this population an understanding of outcomes, particularly patient-centered outcomes, has the potential to facilitate the difficult process of surgical decision making. Patient-centered outcomes such as home time have been reported by patients in other settings to be especially useful for evaluating outcomes of interventions after, for example, stroke.⁹ Although such data are lacking in vascular surgery, we believe the patient-centered outcomes that we have reported, namely, home time, likelihood of discharge to home, risk of discharge to a higher level of care, risk of extended SNF stay, and others, are integral to informed patient decision making.

Our data demonstrate that patients with dementia have a distinctive demographic, comorbidity, and procedural profile compared with patients without dementia. Most important, we showed the strong association of dementia with poor traditional and patient-centered outcomes.

Surgeries in the dementia group had several notable features. First, they were disproportionately urgent/emergent status (60.66% vs 37.93%). Our data do not provide information on patient and provider discussions that could help to clarify the underlying reasons for this observation. It is possible that this may represent provider factors or patient/surrogate reluctance to undergo elective procedures generally. We speculate that a similar desire to decrease risk could account for the lower proportion of high-risk procedures. The surgeries were more likely to be amputations, diagnostic, or related to dialysis or peripheral arterial disease. In contrast, aortic, nonaortic intrathoracic, and intra-abdominal procedures were less common. It is unclear whether this reflects a natural over-representation of certain types of procedures (eg, dialysis access-related procedures often occur serially and therefore would be expected to be frequent) or other phenomena, such as health care preferences in favor of dialysis access interventions, but not more intensive procedures, such as aortic reconstruction.

Outcomes after vascular surgery in the dementia group were poor. After risk adjustment, both 30-day and 90-day mortality rates were worse. The unadjusted ICU admission rate in the dementia group was higher, although this difference disappeared in the multivariable analysis. Although the dementia cohort had a higher unadjusted rate of complications during the index hospitalization, we were surprised to find that this was not different after adjustment, in contrast to the findings of Mehaffey et al.¹⁰ There were several differences that may have contributed to this difference. First, although Mehaffey et al used administrative data, they used their own list rather than one that accords with the chronic conditions warehouse definition. In doing so they omitted codes that we considered essential, such as ICD-9 CM 331.0 for Alzheimer’s disease. Also, our sample size was substantially larger: the study by Mehaffey et al included only 88 patients. These small sample sizes may have been underpowered to detect differences.

Most striking, however, were the uniformly poor patient-centered outcomes. Although patients with dementia are known to have longer hospital LOS,¹¹ the other findings, including the risk of extended SNF admission, discharge to a higher level of care, and home time percentage, are novel and significant findings. These data could be used to help counsel patients facing vascular surgery and appropriately leverage other services, such as palliative care, to help provide health care concordant with patient values and goals.

Strengths of our trial include a large sample size and the use of national data with granular information on patients, procedures, and postoperative trajectories. Weaknesses include a retrospective design and use of Medicare data, which potentially impacts the generalizability of our findings, and the lack of some relevant data points such as dementia severity and anesthesia method. Although dementia severity was unavailable, the use of ICD-9 codes for dementia has been validated in the literature.^12–15 Quan et al,¹² for instance, demonstrated a positive predictive value, negative predictive value, and specificity of 99.2%, 94.6%, and 95.4%, respectively, for the ICD-10 codes for dementia in an audit of more than 4000 charts. van de Vorst et al¹⁵ showed that the positive predictive value was high and not different between men and women and unrelated to type of admission and the patient’s comorbidity burden. Sensitivity, however, has been poor in many studies, and we speculate that that our study may have omitted some patients with dementia, particularly less severe dementia. It is important to note, however, that the inclusion of patients with dementia in the cohort of patients without dementia would bias the results toward the null and therefore the actual differences between the two groups may be even more significant than described here. Importantly, we do not know the outcomes of patients who may have been candidates for vascular surgery but opted for medical treatment. Although our data are from 2011, we find no reason to suspect the findings would be altered with more recent data given the absence of any major advances in care processes for patients with dementia undergoing surgery or with the treatment of dementia itself.

CONCLUSIONS

Fee-for-service Medicare patients with dementia undergoing vascular surgery tend to be older and have more comorbidities compared with vascular patients without a dementia diagnosis. The vascular surgeries are more frequently urgent/emergent status and less likely to be high risk. Procedures are more frequently open and related to dialysis access, peripheral arterial disease, or purely diagnostic. The 30-day and 90-day mortality along with a broad array of patient-centered outcomes were worse in patients with dementia. These include hospital LOS, risk of extended SNF admission, likelihood of discharge home, and home time. These data may be used to help counsel patients with dementia facing vascular surgery.

ARTICLE HIGHLIGHTS.

• Type of Research: Retrospective analysis of Medicare data

• Key Findings: Patients with dementia undergoing vascular surgery had an increased risk of 30-day mortality (odds ratio [OR], 1.21), longer hospital length of stay, extended subacute nursing facility stay (OR, 3.47), were less likely to be discharged home (OR, 0.31), and spent less time at home compared with patients without dementia.

• Take Home Message: Dementia is associated with poor traditional as well as patient-centered outcomes after vascular surgery.